Organization,Availability,Division,SystemType,SystemName,Platform,Model,MlperfModel,Scenario,Result,Accuracy,number_of_nodes,host_processor_model_name,host_processors_per_node,host_processor_core_count,accelerator_model_name,accelerators_per_node,total_accelerators,Location,framework,operating_system,notes,compliance,errors,version,inferred,has_power,Units,weight_data_types,design_power_watts,PrivateID
"AMD","available","closed","datacenter","8xMI350X_2xEPYC_9575F","8xMI350X_2xEPYC_9575F","dlrm-v3","dlrm-v3","Offline","10891.1","DLRM_NE: 0.8677294228203597  DLRM_ACC: 0.6964170290002015  DLRM_AUC: 0.7862871466027813","1","AMD EPYC 9575F","2","64","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/AMD/results/8xMI350X_2xEPYC_9575F/dlrm-v3/Offline/performance/run_1","PyTorch 2.10.0+rocm7.2.3, Triton 3.6.0, ROCm 7.2.3","Ubuntu 24.04 LTS (Noble Numbat), kernel 6.8.0-38-generic","Measured on host smci350-rck-g03-d13-31: 8x AMD Instinct MI350X / 2x AMD EPYC 9575F (64 cores/socket, 2 sockets). Known-good host KMD floor: amdgpu 6.16.6 or newer.. DLRM-v3 ROCm b32 q7,250 GOLD stack (full-causal, C1-off) with degree-5 production gate and TEST08 determinism fixes; container dlrmv3-e2e723 (ROCm 7.2.3).","closed","0","v6.1","0","False","Samples/s","fp16,bf16,fp8","None",""
"AMD","available","closed","datacenter","8xMI350X_2xEPYC_9575F","8xMI350X_2xEPYC_9575F","dlrm-v3","dlrm-v3","Server","9297.78","DLRM_NE: 0.867728327062296  DLRM_ACC: 0.696416918732384  DLRM_AUC: 0.786287540637186","1","AMD EPYC 9575F","2","64","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/AMD/results/8xMI350X_2xEPYC_9575F/dlrm-v3/Server/performance/run_1","PyTorch 2.10.0+rocm7.2.3, Triton 3.6.0, ROCm 7.2.3","Ubuntu 24.04 LTS (Noble Numbat), kernel 6.8.0-38-generic","Measured on host smci350-rck-g03-d13-31: 8x AMD Instinct MI350X / 2x AMD EPYC 9575F (64 cores/socket, 2 sockets). Known-good host KMD floor: amdgpu 6.16.6 or newer.. DLRM-v3 ROCm b32 q7,250 GOLD stack (full-causal, C1-off) with degree-5 production gate and TEST08 determinism fixes; container dlrmv3-e2e723 (ROCm 7.2.3).","closed","0","v6.1","0","False","Queries/s","fp16,bf16,fp8","None",""
"AMD","available","closed","datacenter","Quanta Grand Teton 1.5 MI350","8xMI350X_2xEPYC_9655","wan-2.2-t2v-a14b","text_to_video","Offline","0.0565013","vbench_score: 70.1221","1","AMD EPYC 9655","2","96","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/AMD/results/8xMI350X_2xEPYC_9655/wan-2.2-t2v-a14b/Offline/performance/run_1","xDiT, xfuser 0.4.5, PyTorch 2.9.1+gitff65f5b, ROCm 7.13.0","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Samples/s","bf16, fp8, fp4","None",""
"AMD","available","closed","datacenter","Quanta Grand Teton 1.5 MI350","8xMI350X_2xEPYC_9655","wan-2.2-t2v-a14b","text_to_video","SingleStream","20.603024726","vbench_score: 70.1582","1","AMD EPYC 9655","2","96","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/AMD/results/8xMI350X_2xEPYC_9655/wan-2.2-t2v-a14b/SingleStream/performance/run_1","xDiT, xfuser 0.4.5, PyTorch 2.9.1+gitff65f5b, ROCm 7.13.0","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Latency (s)","bf16, fp8, fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","dlrm-v3","dlrm-v3","Offline","13027.7","DLRM_NE: 0.8681176645656181  DLRM_ACC: 0.6963383089448092  DLRM_AUC: 0.7862169525257562","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/dlrm-v3/Offline/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Samples/s","fp16,bf16,fp8","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","dlrm-v3","dlrm-v3","Server","12198.8","DLRM_NE: 0.8681164332983042  DLRM_ACC: 0.696331698458935  DLRM_AUC: 0.786212286822123","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/dlrm-v3/Server/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Queries/s","fp16,bf16,fp8","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","gpt-oss-120b","gpt-oss-120b","Offline","121818.0","exact_match: 83.706","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/gpt-oss-120b/Offline/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","gpt-oss-120b","gpt-oss-120b","Server","113241.0","exact_match: 83.451","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/gpt-oss-120b/Server/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","llama2-70b-99.9","llama2-70b-99.9","Interactive","73191.5","ROUGE1: 44.5125  ROUGE2: 22.1593  ROUGEL: 28.9205  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/llama2-70b-99.9/Interactive/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","llama2-70b-99.9","llama2-70b-99.9","Offline","106517.0","ROUGE1: 44.5481  ROUGE2: 22.1846  ROUGEL: 28.9482  TOKENS_PER_SAMPLE: 283.2","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/llama2-70b-99.9/Offline/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","llama2-70b-99.9","llama2-70b-99.9","Server","103275.0","ROUGE1: 44.5266  ROUGE2: 22.1733  ROUGEL: 28.9289  TOKENS_PER_SAMPLE: 284.2","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/llama2-70b-99.9/Server/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","wan-2.2-t2v-a14b","text_to_video","Offline","0.0731282","vbench_score: 69.8899","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/wan-2.2-t2v-a14b/Offline/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Samples/s","bf16, fp8, fp4","None",""
"AMD","available","closed","datacenter","smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","8xMI355X_2xEPYC_9575F","wan-2.2-t2v-a14b","text_to_video","SingleStream","16.124484197","vbench_score: 69.9691","1","AMD EPYC 9575F","8","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/AMD/results/8xMI355X_2xEPYC_9575F/wan-2.2-t2v-a14b/SingleStream/performance/run_1","PyTorch 2.10.0+git8514f05, ROCm 7.2.2","Ubuntu 22.04.5 LTS","","closed","0","v6.1","0","False","Latency (s)","bf16, fp8, fp4","None",""
"AMD","available","closed","datacenter","9x8xMI355X_2xEPYC_9575F","9x8xMI355X_2xEPYC_9575F","gpt-oss-120b","gpt-oss-120b","Offline","1039900.0","exact_match: 83.230","9","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","72","./closed/AMD/results/9x8xMI355X_2xEPYC_9575F/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.1.dev0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","9 nodes x 8 AMD Instinct MI355X (gfx950) = 72 GPUs, 288GB HBM3E each; distributed asynchronous ZMQ backend. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization; 72-GPU distributed ZMQ (async server / sync offline)","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"AMD","available","closed","datacenter","9x8xMI355X_2xEPYC_9575F","9x8xMI355X_2xEPYC_9575F","gpt-oss-120b","gpt-oss-120b","Server","964468.0","exact_match: 83.446","9","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","72","./closed/AMD/results/9x8xMI355X_2xEPYC_9575F/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.1.dev0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","9 nodes x 8 AMD Instinct MI355X (gfx950) = 72 GPUs, 288GB HBM3E each; distributed asynchronous ZMQ backend. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization; 72-GPU distributed ZMQ (async server / sync offline)","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"ASUSTeK","available","closed","datacenter","XA NB3I-E12","XA_NB3I-E12_B300x8_TRT","deepseek-r1","deepseek-r1","Offline","69846.9","exact_match: 80.76572470373746  TOKENS_PER_SAMPLE: 3759.735870556062","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/ASUSTeK/results/XA_NB3I-E12_B300x8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"ASUSTeK","available","closed","datacenter","XA NB3I-E12","XA_NB3I-E12_B300x8_TRT","deepseek-r1","deepseek-r1","Server","54568.6","exact_match: 80.9024612579763  TOKENS_PER_SAMPLE: 3802.827256153145","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/ASUSTeK/results/XA_NB3I-E12_B300x8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"ASUSTeK","available","closed","datacenter","XA NB3I-E12","XA_NB3I-E12_B300x8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","109368.0","exact_match: 83.572","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/ASUSTeK/results/XA_NB3I-E12_B300x8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"ASUSTeK","available","closed","datacenter","XA NB3I-E12","XA_NB3I-E12_B300x8_TRT","gpt-oss-120b","gpt-oss-120b","Server","109949.0","exact_match: 83.403","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/ASUSTeK/results/XA_NB3I-E12_B300x8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"ASUSTeK","available","closed","datacenter","XA NB3I-E12","XA_NB3I-E12_B300x8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","113007.0","ROUGE1: 44.7974  ROUGE2: 22.3739  ROUGEL: 29.1541  TOKENS_PER_SAMPLE: 272.7","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/ASUSTeK/results/XA_NB3I-E12_B300x8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"ASUSTeK","available","closed","datacenter","XA NB3I-E12","XA_NB3I-E12_B300x8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","113455.0","ROUGE1: 44.8853  ROUGE2: 22.409  ROUGEL: 29.2169  TOKENS_PER_SAMPLE: 272.2","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/ASUSTeK/results/XA_NB3I-E12_B300x8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Atlas_Inference","available","closed","edge","NVIDIA DGX Spark (GB10 Grace Blackwell Superchip)","DGX_Spark_GB10","qwen3.6-27b","agentic edge","SingleStream","3807.66351136147","0.8723618090452261","1","NVIDIA Grace (20-core Arm v9.2)","1","20","NVIDIA GB10","1","1","./closed/Atlas_Inference/results/DGX_Spark_GB10/qwen3.6-27b/SingleStream/performance/run_1","Atlas (spark), CUDA NVFP4","Ubuntu 24.04.4 LTS","NVIDIA DGX Spark GB10 Grace Blackwell Superchip with 128 GB unified LPDDR5X memory (~273 GB/s) shared between the Grace CPU and the GB10 GPU; roughly 119-121 GiB of that is reportable to the OS as usable. GB10 GPU is Blackwell, sm_121a / compute capability 12.1.. Atlas NVFP4 serve; chain-widened MTP verify K=4 (3 drafts) + drafter context prefill/carry; GDN register-resident warm-replay (ATLAS_GDN_REGRESIDENT, merged default-on upstream as 063d87ee); see results/.../README.md.","closed","0","v6.1","1","False","Latency (ms)","nvfp4 (E2M1 4-bit + FP8 block scales), GDN included","None",""
"Atlas_Inference","available","closed","edge","AMD Strix Halo desktop (Ryzen AI Max+ 395 / Radeon 8060S)","Strix_Halo","qwen3.6-27b","agentic edge","SingleStream","7058.989199685204","0.8623115577889447","1","AMD Ryzen AI Max+ 395","1","16","AMD Radeon 8060S Graphics","1","1","./closed/Atlas_Inference/results/Strix_Halo/qwen3.6-27b/SingleStream/performance/run_1","Atlas (spark), native-HIP NVFP4","Ubuntu 24.04 (Linux), ROCm 7.13","AMD Strix Halo APU (Ryzen AI Max+ 395) with 128 GB unified LPDDR5X memory shared between the CPU and the integrated GPU; ~62.5 GB of that is GPU-addressable via GTT. Radeon 8060S Graphics is gfx1151, RDNA3.5.. Atlas NVFP4 native-HIP serve; MTP K=3 (2 drafts) + drafter context-prefill; see results/.../README.md.","closed","0","v6.1","1","False","Latency (ms)","nvfp4 (E2M1 4-bit + FP8 block scales)","None",""
"Azure","available","closed","datacenter","Azure GB200 NVL72 (288x GB200-186GB_aarch64, TensorRT)","Azure-GB200-NVL72_186GBx288","deepseek-r1","deepseek-r1","Offline","2091190.0","exact_match: 81.56335460346399  TOKENS_PER_SAMPLE: 3758.743163172288","72","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","288","./closed/Azure/results/Azure-GB200-NVL72_186GBx288/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Azure","available","closed","datacenter","Azure GB200 NVL72 (288x GB200-186GB_aarch64, TensorRT)","Azure-GB200-NVL72_186GBx288","deepseek-r1","deepseek-r1","Server","1598510.0","exact_match: 80.85688240656336  TOKENS_PER_SAMPLE: 3746.623518687329","72","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","288","./closed/Azure/results/Azure-GB200-NVL72_186GBx288/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Azure","available","closed","datacenter","Azure GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","Azure-GB200-NVL72_186GBx72","deepseek-r1","deepseek-r1","Offline","542075.0","exact_match: 81.35824977210574  TOKENS_PER_SAMPLE: 3739.669553327256","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/Azure/results/Azure-GB200-NVL72_186GBx72/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Azure","available","closed","datacenter","Azure GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","Azure-GB200-NVL72_186GBx72","deepseek-r1","deepseek-r1","Server","426796.0","exact_match: 81.51777575205105  TOKENS_PER_SAMPLE: 3732.908158614403","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/Azure/results/Azure-GB200-NVL72_186GBx72/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Azure","available","closed","datacenter","Azure GB300 NVL72 (72x GB300-288GB_aarch64 TensorRT)","Azure-GB300-NVL72_288GBx72","deepseek-r1","deepseek-r1","Interactive","260098.0","exact_match: 80.81130355515042  TOKENS_PER_SAMPLE: 3720.641750227894","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Azure/results/Azure-GB300-NVL72_288GBx72/deepseek-r1/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Azure","available","closed","datacenter","Azure GB300 NVL72 (72x GB300-288GB_aarch64 TensorRT)","Azure-GB300-NVL72_288GBx72","deepseek-r1","deepseek-r1","Offline","661143.0","exact_match: 81.47219690063811  TOKENS_PER_SAMPLE: 3779.8644029170464","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Azure/results/Azure-GB300-NVL72_288GBx72/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Azure","available","closed","datacenter","Azure GB300 NVL72 (72x GB300-288GB_aarch64 TensorRT)","Azure-GB300-NVL72_288GBx72","deepseek-r1","deepseek-r1","Server","589391.0","exact_match: 81.13035551504102  TOKENS_PER_SAMPLE: 3771.7791704649044","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Azure/results/Azure-GB300-NVL72_288GBx72/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C240M8_RTXPro4500_32GBx5_TRT","llama3.1-8b","llama3.1-8b","Interactive","8494.68","ROUGE1: 38.7733  ROUGE2: 16.0169  ROUGEL: 24.5358  ROUGELSUM: 35.6947  GEN_LEN: 8179270","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA RTX PRO 4500 Blackwell","5","5","./closed/Cisco/results/C240M8_RTXPro4500_32GBx5_TRT/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C240M8_RTXPro4500_32GBx5_TRT","llama3.1-8b","llama3.1-8b","Offline","12017.7","ROUGE1: 38.7698  ROUGE2: 16.0132  ROUGEL: 24.5304  ROUGELSUM: 35.6924  GEN_LEN: 8178850","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA RTX PRO 4500 Blackwell","5","5","./closed/Cisco/results/C240M8_RTXPro4500_32GBx5_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C240M8_RTXPro4500_32GBx5_TRT","llama3.1-8b","llama3.1-8b","Server","11426.1","ROUGE1: 38.7615  ROUGE2: 16.0056  ROUGEL: 24.5235  ROUGELSUM: 35.6866  GEN_LEN: 8178120","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA RTX PRO 4500 Blackwell","5","5","./closed/Cisco/results/C240M8_RTXPro4500_32GBx5_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C240M8_RTXPro4500_32GBx5_TRT","whisper","whisper","Offline","4808.98","ACCURACY: 97.83682442399383","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA RTX PRO 4500 Blackwell","5","5","./closed/Cisco/results/C240M8_RTXPro4500_32GBx5_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C845A_M8_RTXPro4500_32GBx8_TRT","llama3.1-8b","llama3.1-8b","Interactive","14316.2","ROUGE1: 38.7705  ROUGE2: 16.0134  ROUGEL: 24.5326  ROUGELSUM: 35.6934  GEN_LEN: 8179270","1","AMD EPYC 9655","2","96","NVIDIA RTX PRO 4500 Blackwell","8","8","./closed/Cisco/results/C845A_M8_RTXPro4500_32GBx8_TRT/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C845A_M8_RTXPro4500_32GBx8_TRT","llama3.1-8b","llama3.1-8b","Offline","18800.9","ROUGE1: 38.7819  ROUGE2: 16.0194  ROUGEL: 24.5377  ROUGELSUM: 35.7086  GEN_LEN: 8178827","1","AMD EPYC 9655","2","96","NVIDIA RTX PRO 4500 Blackwell","8","8","./closed/Cisco/results/C845A_M8_RTXPro4500_32GBx8_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C845A_M8_RTXPro4500_32GBx8_TRT","llama3.1-8b","llama3.1-8b","Server","18186.2","ROUGE1: 38.7633  ROUGE2: 16.0118  ROUGEL: 24.5232  ROUGELSUM: 35.6834  GEN_LEN: 8178277","1","AMD EPYC 9655","2","96","NVIDIA RTX PRO 4500 Blackwell","8","8","./closed/Cisco/results/C845A_M8_RTXPro4500_32GBx8_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","C845A_M8_RTXPro4500_32GBx8_TRT","whisper","whisper","Offline","7777.66","ACCURACY: 97.82825927653055","1","AMD EPYC 9655","2","96","NVIDIA RTX PRO 4500 Blackwell","8","8","./closed/Cisco/results/C845A_M8_RTXPro4500_32GBx8_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","15508.4","exact_match: 83.570","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","14817.4","exact_match: 84.117","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","29904.0","ROUGE1: 44.7661  ROUGE2: 22.3346  ROUGEL: 29.1319  TOKENS_PER_SAMPLE: 274.4","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","29316.4","ROUGE1: 44.7625  ROUGE2: 22.3336  ROUGEL: 29.1052  TOKENS_PER_SAMPLE: 273.5","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","llama3.1-8b","llama3.1-8b","Interactive","46164.4","ROUGE1: 38.7602  ROUGE2: 15.9962  ROUGEL: 24.5347  ROUGELSUM: 35.6851  GEN_LEN: 8180275","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","llama3.1-8b","llama3.1-8b","Offline","50453.7","ROUGE1: 38.769  ROUGE2: 16.0147  ROUGEL: 24.5312  ROUGELSUM: 35.6942  GEN_LEN: 8178808","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","llama3.1-8b","llama3.1-8b","Server","49174.7","ROUGE1: 38.7598  ROUGE2: 16.0015  ROUGEL: 24.5351  ROUGELSUM: 35.6872  GEN_LEN: 8179087","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","C845A_M8_RTXPro6000_96GBx8_TRT","whisper","whisper","Offline","18991.8","ACCURACY: 97.8425345223027","1","AMD EPYC 9575F","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Cisco/results/C845A_M8_RTXPro6000_96GBx8_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","70325.4","exact_match: 80.92525068368278  TOKENS_PER_SAMPLE: 3782.986326344576","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","deepseek-r1","deepseek-r1","Server","66171.6","exact_match: 80.76572470373746  TOKENS_PER_SAMPLE: 3811.0783956244304","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","llama3.1-8b","llama3.1-8b","Interactive","129168.0","ROUGE1: 38.4828  ROUGE2: 15.9176  ROUGEL: 24.3231  ROUGELSUM: 35.6535  GEN_LEN: 8186168","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","llama3.1-8b","llama3.1-8b","Offline","171114.0","ROUGE1: 38.4898  ROUGE2: 15.9329  ROUGEL: 24.3413  ROUGELSUM: 35.6705  GEN_LEN: 8186559","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","llama3.1-8b","llama3.1-8b","Server","173950.0","ROUGE1: 38.4912  ROUGE2: 15.9156  ROUGEL: 24.3266  ROUGELSUM: 35.6593  GEN_LEN: 8186471","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.06127496829802798","0.7028246725986534","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","20.004214518034722","0.6994130989447821","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"Cisco","available","closed","datacenter","Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","C880A_M8_B300_270GBx8_TRT","whisper","whisper","Offline","52221.8","ACCURACY: 97.8263559104276","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Cisco/results/C880A_M8_B300_270GBx8_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Cisco","available","closed","datacenter","Cisco_UCS_B300x16_G200_4x2_RO","Cisco_UCS_B300x16_G200_4x2_RO","deepseek-r1","deepseek-r1","Offline","137067.0","exact_match: 81.33546034639927  TOKENS_PER_SAMPLE: 3825.4619416590704","2","Intel(R) Xeon(R) 6776P","2","128","NVIDIA B300-SXM-270GB","8","16","./closed/Cisco/results/Cisco_UCS_B300x16_G200_4x2_RO/deepseek-r1/Offline/performance/run_1","TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.17.0-35-generic","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco_UCS_B300x16_G200_4x2_RO","Cisco_UCS_B300x16_G200_4x2_RO","deepseek-r1","deepseek-r1","Server","88915.5","exact_match: 81.28988149498633  TOKENS_PER_SAMPLE: 3794.9457611668186","2","Intel(R) Xeon(R) 6776P","2","128","NVIDIA B300-SXM-270GB","8","16","./closed/Cisco/results/Cisco_UCS_B300x16_G200_4x2_RO/deepseek-r1/Server/performance/run_1","TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.17.0-35-generic","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco_UCS_B300x16_G200_4x2_RO","Cisco_UCS_B300x16_G200_4x2_RO","gpt-oss-120b","gpt-oss-120b","Offline","220527.0","exact_match: 83.296","2","Intel(R) Xeon(R) 6776P","2","128","NVIDIA B300-SXM-270GB","8","16","./closed/Cisco/results/Cisco_UCS_B300x16_G200_4x2_RO/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.17.0-35-generic","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco_UCS_B300x16_G200_4x2_RO","Cisco_UCS_B300x16_G200_4x2_RO","gpt-oss-120b","gpt-oss-120b","Server","214774.0","exact_match: 83.572","2","Intel(R) Xeon(R) 6776P","2","128","NVIDIA B300-SXM-270GB","8","16","./closed/Cisco/results/Cisco_UCS_B300x16_G200_4x2_RO/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.17.0-35-generic","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco_UCS_B300x16_G200_4x2_RO","Cisco_UCS_B300x16_G200_4x2_RO","llama2-70b-99.9","llama2-70b-99.9","Offline","227570.0","ROUGE1: 44.8082  ROUGE2: 22.3989  ROUGEL: 29.1755  TOKENS_PER_SAMPLE: 272.1","2","Intel(R) Xeon(R) 6776P","2","128","NVIDIA B300-SXM-270GB","8","16","./closed/Cisco/results/Cisco_UCS_B300x16_G200_4x2_RO/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.17.0-35-generic","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco_UCS_B300x16_G200_4x2_RO","Cisco_UCS_B300x16_G200_4x2_RO","llama2-70b-99.9","llama2-70b-99.9","Server","179597.0","ROUGE1: 44.7969  ROUGE2: 22.3691  ROUGEL: 29.184  TOKENS_PER_SAMPLE: 272.7","2","Intel(R) Xeon(R) 6776P","2","128","NVIDIA B300-SXM-270GB","8","16","./closed/Cisco/results/Cisco_UCS_B300x16_G200_4x2_RO/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.17.0-35-generic","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","gpt-oss-120b","gpt-oss-120b","Offline","94249.3","exact_match: 83.57105741751104","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/gpt-oss-120b/Offline/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","gpt-oss-120b","gpt-oss-120b","Server","73357.6","exact_match: 83.43465835281891","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/gpt-oss-120b/Server/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","llama2-70b-99.9","llama2-70b-99.9","Interactive","70354.5","ROUGE1: 44.5557  ROUGE2: 22.154  ROUGEL: 28.907  TOKENS_PER_SAMPLE: 283.6","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/llama2-70b-99.9/Interactive/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","1","False","Tokens/s","fp8,fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","llama2-70b-99.9","llama2-70b-99.9","Offline","99683.8","ROUGE1: 44.4169  ROUGE2: 22.068  ROUGEL: 28.7957  TOKENS_PER_SAMPLE: 286.6","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/llama2-70b-99.9/Offline/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp8,fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","llama2-70b-99.9","llama2-70b-99.9","Server","95104.6","ROUGE1: 44.6044  ROUGE2: 22.1926  ROUGEL: 28.9466  TOKENS_PER_SAMPLE: 284.7","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/llama2-70b-99.9/Server/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp8,fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","llama3.1-8b","llama3.1-8b","Interactive","121557.0","ROUGE1: 38.6105  ROUGE2: 16.0699  ROUGEL: 24.5048  ROUGELSUM: 35.5444  GEN_LEN: 8148492","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/llama3.1-8b/Interactive/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","1","False","Tokens/s","fp8,fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","llama3.1-8b","llama3.1-8b","Offline","163292.0","ROUGE1: 38.5675  ROUGE2: 16.0757  ROUGEL: 24.5063  ROUGELSUM: 35.5154  GEN_LEN: 8141814","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/llama3.1-8b/Offline/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp8,fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR","llama3.1-8b","llama3.1-8b","Server","161110.0","ROUGE1: 38.6437  ROUGE2: 16.1045  ROUGEL: 24.5278  ROUGELSUM: 35.5601  GEN_LEN: 8147699","2","AMD EPYC","2","128","NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","8","16","./closed/Cisco/results/Cisco_UCS_H200x8_MI350Xx8_G200_2x2_TOR/llama3.1-8b/Server/performance/run_1","vLLM","Ubuntu 22.04","","closed","0","v6.1","0","False","Tokens/s","fp8,fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","X210M8_X580P_RTXPro4500_32GBx4_TRT","llama3.1-8b","llama3.1-8b","Interactive","6883.94","ROUGE1: 38.7575  ROUGE2: 16.0054  ROUGEL: 24.5233  ROUGELSUM: 35.6829  GEN_LEN: 8179124","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 4500 Blackwell","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro4500_32GBx4_TRT/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","X210M8_X580P_RTXPro4500_32GBx4_TRT","llama3.1-8b","llama3.1-8b","Offline","9518.66","ROUGE1: 38.7664  ROUGE2: 16.0107  ROUGEL: 24.5281  ROUGELSUM: 35.6917  GEN_LEN: 8178365","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 4500 Blackwell","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro4500_32GBx4_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","X210M8_X580P_RTXPro4500_32GBx4_TRT","llama3.1-8b","llama3.1-8b","Server","9277.73","ROUGE1: 38.7836  ROUGE2: 16.0298  ROUGEL: 24.5367  ROUGELSUM: 35.7048  GEN_LEN: 8178628","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 4500 Blackwell","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro4500_32GBx4_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","X210M8_X580P_RTXPro4500_32GBx4_TRT","whisper","whisper","Offline","3877.67","ACCURACY: 97.86822996469256","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 4500 Blackwell","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro4500_32GBx4_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X210M8_X580P_RTXPro6000_96GBx4_TRT","gpt-oss-120b","gpt-oss-120b","Offline","7801.36","exact_match: 83.969","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro6000_96GBx4_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X210M8_X580P_RTXPro6000_96GBx4_TRT","gpt-oss-120b","gpt-oss-120b","Server","6950.76","exact_match: 83.637","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro6000_96GBx4_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X210M8_X580P_RTXPro6000_96GBx4_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","14860.7","ROUGE1: 44.7766  ROUGE2: 22.3407  ROUGEL: 29.1379  TOKENS_PER_SAMPLE: 275.2","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro6000_96GBx4_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X210M8_X580P_RTXPro6000_96GBx4_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","14333.3","ROUGE1: 44.7151  ROUGE2: 22.3389  ROUGEL: 29.0874  TOKENS_PER_SAMPLE: 274.6","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X210M8_X580P_RTXPro6000_96GBx4_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X215M8_X580P_RTXPro6000_96GBx4_TRT","llama3.1-8b","llama3.1-8b","Interactive","22447.0","ROUGE1: 38.7706  ROUGE2: 16.012  ROUGEL: 24.5423  ROUGELSUM: 35.699  GEN_LEN: 8179184","1","AMD EPYC 9845","2","160","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X215M8_X580P_RTXPro6000_96GBx4_TRT/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X215M8_X580P_RTXPro6000_96GBx4_TRT","llama3.1-8b","llama3.1-8b","Offline","24913.5","ROUGE1: 38.7701  ROUGE2: 16.0155  ROUGEL: 24.534  ROUGELSUM: 35.6954  GEN_LEN: 8178898","1","AMD EPYC 9845","2","160","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X215M8_X580P_RTXPro6000_96GBx4_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X215M8_X580P_RTXPro6000_96GBx4_TRT","llama3.1-8b","llama3.1-8b","Server","24498.6","ROUGE1: 38.766  ROUGE2: 16.0082  ROUGEL: 24.5262  ROUGELSUM: 35.6898  GEN_LEN: 8179590","1","AMD EPYC 9845","2","160","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X215M8_X580P_RTXPro6000_96GBx4_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Cisco","available","closed","datacenter","Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","X215M8_X580P_RTXPro6000_96GBx4_TRT","whisper","whisper","Offline","9236.94","ACCURACY: 97.8339693748394","1","AMD EPYC 9845","2","160","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/Cisco/results/X215M8_X580P_RTXPro6000_96GBx4_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Cisco","available","closed","datacenter","X410M8_1_node_4S_GNR_86C","X410M8_1_node_4S_GNR_86C","llama3_1-8b","llama3.1-8b","Offline","2068.63","ROUGE1: 38.6284  ROUGE2: 15.804  ROUGEL: 24.4033  ROUGELSUM: 35.6616  GEN_LEN: 8196513","1","Intel(R) Xeon(R) 6788P","4","86","N/A","0","0","./closed/Cisco/results/X410M8_1_node_4S_GNR_86C/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Cisco","available","closed","datacenter","X410M8_1_node_4S_GNR_86C","X410M8_1_node_4S_GNR_86C","llama3_1-8b","llama3.1-8b","Server","1030.45","ROUGE1: 38.8382  ROUGE2: 16.0493  ROUGEL: 24.6331  ROUGELSUM: 35.8769  GEN_LEN: 8171982","1","Intel(R) Xeon(R) 6788P","4","86","N/A","0","0","./closed/Cisco/results/X410M8_1_node_4S_GNR_86C/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04.3","","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Cisco","available","closed","datacenter","X410M8_1_node_4S_GNR_86C","X410M8_1_node_4S_GNR_86C","whisper","whisper","Offline","2970.57","ACCURACY: 97.94665880091725","1","Intel(R) Xeon(R) 6788P","4","86","N/A","0","0","./closed/Cisco/results/X410M8_1_node_4S_GNR_86C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B200 SXM (8x B200-SXM-180GB)","B200-SXM-180GBx8_TRT","deepseek-r1","deepseek-r1","Offline","58913.2","exact_match: 80.9024612579763  TOKENS_PER_SAMPLE: 3888.048541476755","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA B200-SXM-180GB","8","8","./closed/CoreWeave/results/B200-SXM-180GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B200 SXM (8x B200-SXM-180GB)","B200-SXM-180GBx8_TRT","deepseek-r1","deepseek-r1","Server","56121.3","exact_match: 81.60893345487693  TOKENS_PER_SAMPLE: 3816.3475387420235","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA B200-SXM-180GB","8","8","./closed/CoreWeave/results/B200-SXM-180GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B200 SXM (8x B200-SXM-180GB)","B200-SXM-180GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","91487.4","exact_match: 83.901","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA B200-SXM-180GB","8","8","./closed/CoreWeave/results/B200-SXM-180GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B200 SXM (8x B200-SXM-180GB)","B200-SXM-180GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","90204.7","exact_match: 83.728","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA B200-SXM-180GB","8","8","./closed/CoreWeave/results/B200-SXM-180GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B200 SXM (8x B200-SXM-180GB)","B200-SXM-180GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","102703.0","ROUGE1: 44.833  ROUGE2: 22.4096  ROUGEL: 29.2097  TOKENS_PER_SAMPLE: 271.6","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA B200-SXM-180GB","8","8","./closed/CoreWeave/results/B200-SXM-180GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B200 SXM (8x B200-SXM-180GB)","B200-SXM-180GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","102398.0","ROUGE1: 44.8305  ROUGE2: 22.4096  ROUGEL: 29.21  TOKENS_PER_SAMPLE: 271.5","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA B200-SXM-180GB","8","8","./closed/CoreWeave/results/B200-SXM-180GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B300 SXM (8x B300-SXM-270GB)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","112840.0","exact_match: 83.290","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/CoreWeave/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B300 SXM (8x B300-SXM-270GB)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","111247.0","exact_match: 83.467","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/CoreWeave/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B300 SXM (8x B300-SXM-270GB)","B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","115530.0","ROUGE1: 44.7888  ROUGE2: 22.3737  ROUGEL: 29.1521  TOKENS_PER_SAMPLE: 272.9","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/CoreWeave/results/B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave B300 SXM (8x B300-SXM-270GB)","B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","114576.0","ROUGE1: 44.8422  ROUGE2: 22.4119  ROUGEL: 29.1975  TOKENS_PER_SAMPLE: 272.8","1","Intel(R) Xeon(R) 6747P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/CoreWeave/results/B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","529991.0","exact_match: 80.76572470373746  TOKENS_PER_SAMPLE: 3815.436645396536","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB200","4","72","./closed/CoreWeave/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","GB200 TGP 1200W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","419778.0","exact_match: 81.03919781221514  TOKENS_PER_SAMPLE: 3802.557201458523","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB200","4","72","./closed/CoreWeave/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","GB200 TGP 1200W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Offline","911566.0","exact_match: 83.394","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB200","4","72","./closed/CoreWeave/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","GB200 TGP 1200W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Server","901058.0","exact_match: 83.424","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB200","4","72","./closed/CoreWeave/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","GB200 TGP 1200W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","894002.0","ROUGE1: 44.8296  ROUGE2: 22.4099  ROUGEL: 29.2105  TOKENS_PER_SAMPLE: 271.5","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB200","4","72","./closed/CoreWeave/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","GB200 TGP 1200W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","910621.0","ROUGE1: 44.8247  ROUGE2: 22.3971  ROUGEL: 29.2015  TOKENS_PER_SAMPLE: 271.7","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB200","4","72","./closed/CoreWeave/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","GB200 TGP 1200W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (4x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Offline","72.59800294104302","0.7873886028433235","1","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","4","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","vLLM CentML:mlperf-inf-mm-q3vl-v6.0, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","Single 4-GPU GB300 node, TGP 1400W per GPU","closed","0","v6.1","1","False","Samples/s","fp4 weights, fp8 KV cache","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (4x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Server","64.13202030676248","0.7880294101448477","1","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","4","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","vLLM CentML:mlperf-inf-mm-q3vl-v6.0, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0","Ubuntu 24.04","Single 4-GPU GB300 node, TGP 1400W per GPU","closed","0","v6.1","1","False","Queries/s","fp4 weights, fp8 KV cache","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Offline","1197710.0","exact_match: 83.098","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","72","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","GB300 TGP 1400W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Server","1160490.0","exact_match: 82.875","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","72","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","GB300 TGP 1400W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","1136100.0","ROUGE1: 44.8147  ROUGE2: 22.3736  ROUGEL: 29.189  TOKENS_PER_SAMPLE: 272.4","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","72","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","GB300 TGP 1400W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","944902.0","ROUGE1: 44.8524  ROUGE2: 22.4096  ROUGEL: 29.218  TOKENS_PER_SAMPLE: 272.3","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","72","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","GB300 TGP 1400W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Offline","1135.2500157943905","0.78717027204421","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","72","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","GB300 TGP 1400W","closed","0","v6.1","1","False","Samples/s","fp4 weights, fp8 KV cache","None",""
"CoreWeave","available","closed","datacenter","CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Server","1195.5907586337776","0.7875410543420632","18","NVIDIA Grace (Neoverse-V2)","2","72","NVIDIA GB300","4","72","./closed/CoreWeave/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","GB300 TGP 1400W","closed","0","v6.1","1","False","Queries/s","fp4 weights, fp8 KV cache","None",""
"Crusoe","available","closed","datacenter","Crusoe Cloud 64x8 MI355X (512 GPU) — DeepSeek-R1 / SGLang","512xMI355X_Crusoe_DS","deepseek-r1","deepseek-r1","Offline","2901950.0","exact_match: 80.69735642661804  TOKENS_PER_SAMPLE: 3898.373746581586","64","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","512","./closed/Crusoe/results/512xMI355X_Crusoe_DS/deepseek-r1/Offline/performance/run_1","SGLang 0.5.15.post1 (ROCm 7.2.0), PyTorch 2.x","Ubuntu 24.04 (node amdgpu driver 6.14.x; container ROCm 7.2.0 userspace)","Crusoe Cloud mi355x-288gb-roce.8x, 64 nodes = 512 MI355X GPUs. First-ever 512-GPU DeepSeek-R1 (Offline + Server).. AMD MLPerf Inference v6.1 DeepSeek-R1 harness (SGLang backend). Model amd/Deepseek-S3_sq_a05_v2_mlperf6_1 — MXFP4 experts / FP8 KV cache / BF16 MLA-absorb. Per node TP8 / EP8 / DP-attention8 / PP1 (one 8-GPU replica); device_count=512. Server uses engine warmup (dsr_code6_1 WarmUp.ENCODED_SAMPLES) + prefill-delayer for TTFT/TPOT.","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Crusoe","available","closed","datacenter","Crusoe Cloud 64x8 MI355X (512 GPU) — DeepSeek-R1 / SGLang","512xMI355X_Crusoe_DS","deepseek-r1","deepseek-r1","Server","2405310.0","exact_match: 80.69735642661804  TOKENS_PER_SAMPLE: 3898.373746581586","64","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","512","./closed/Crusoe/results/512xMI355X_Crusoe_DS/deepseek-r1/Server/performance/run_1","SGLang 0.5.15.post1 (ROCm 7.2.0), PyTorch 2.x","Ubuntu 24.04 (node amdgpu driver 6.14.x; container ROCm 7.2.0 userspace)","Crusoe Cloud mi355x-288gb-roce.8x, 64 nodes = 512 MI355X GPUs. First-ever 512-GPU DeepSeek-R1 (Offline + Server).. AMD MLPerf Inference v6.1 DeepSeek-R1 harness (SGLang backend). Model amd/Deepseek-S3_sq_a05_v2_mlperf6_1 — MXFP4 experts / FP8 KV cache / BF16 MLA-absorb. Per node TP8 / EP8 / DP-attention8 / PP1 (one 8-GPU replica); device_count=512. Server uses engine warmup (dsr_code6_1 WarmUp.ENCODED_SAMPLES) + prefill-delayer for TTFT/TPOT.","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Crusoe","available","closed","datacenter","Crusoe Cloud 64x8 MI355X (512 GPU) — gpt-oss-120b / vLLM","512xMI355X_Crusoe_GPT","gpt-oss-120b","gpt-oss-120b","Offline","5749440.0","exact_match: 83.706","64","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","512","./closed/Crusoe/results/512xMI355X_Crusoe_GPT/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.1 (ROCm 7.2.2), PyTorch 2.10","Ubuntu 24.04 (node amdgpu driver 6.14.x; container ROCm 7.2.2 userspace)","Crusoe Cloud mi355x-288gb-roce.8x, 64 nodes = 512 MI355X GPUs. First-ever 512-GPU gpt-oss-120b (Offline + Server).. AMD MLPerf Inference v6.1 gpt-oss-120b harness (vLLM backend). Native-MXFP4 model openai/gpt-oss-120b@b5c939d (NOT the v6.0 Quark repack). tensor_parallel_size=1 (one GPU per replica, 8 replicas/node); device_count=512. Dedicated ZMQ head dispatches Offline sample-ranges / Server per-query and relays token streams.","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"Crusoe","available","closed","datacenter","Crusoe Cloud 64x8 MI355X (512 GPU) — gpt-oss-120b / vLLM","512xMI355X_Crusoe_GPT","gpt-oss-120b","gpt-oss-120b","Server","5392930.0","exact_match: 83.706","64","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","512","./closed/Crusoe/results/512xMI355X_Crusoe_GPT/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.1 (ROCm 7.2.2), PyTorch 2.10","Ubuntu 24.04 (node amdgpu driver 6.14.x; container ROCm 7.2.2 userspace)","Crusoe Cloud mi355x-288gb-roce.8x, 64 nodes = 512 MI355X GPUs. First-ever 512-GPU gpt-oss-120b (Offline + Server).. AMD MLPerf Inference v6.1 gpt-oss-120b harness (vLLM backend). Native-MXFP4 model openai/gpt-oss-120b@b5c939d (NOT the v6.0 Quark repack). tensor_parallel_size=1 (one GPU per replica, 8 replicas/node); device_count=512. Dedicated ZMQ head dispatches Offline sample-ranges / Server per-query and relays token streams.","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (16x GB200-186GB_aarch64, Dynamo)","GB200-NVL72_GB200-186GB_aarch64x16_Dynamo","qwen3-vl-235b-a22b","VLM","Interactive","39.85079133425917","0.7860671731639475","4","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","16","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x16_Dynamo/qwen3-vl-235b-a22b/Interactive/performance/run_1","NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4, FP8 (KV cache)","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (16x GB200-186GB_aarch64, Dynamo)","GB200-NVL72_GB200-186GB_aarch64x16_Dynamo","qwen3-vl-235b-a22b","VLM","Offline","230.088792881784","0.7880099213678822","4","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","16","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x16_Dynamo/qwen3-vl-235b-a22b/Offline/performance/run_1","NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","NVFP4, FP8 (KV cache)","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (16x GB200-186GB_aarch64, Dynamo)","GB200-NVL72_GB200-186GB_aarch64x16_Dynamo","qwen3-vl-235b-a22b","VLM","Server","198.9444574056901","0.7881503918126641","4","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","16","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x16_Dynamo/qwen3-vl-235b-a22b/Server/performance/run_1","NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4, FP8 (KV cache)","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, Dynamo)","GB200-NVL72_GB200-186GB_aarch64x4_Dynamo","qwen3-vl-235b-a22b","VLM","Interactive","3.9940567447504782","0.785571974099286","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x4_Dynamo/qwen3-vl-235b-a22b/Interactive/performance/run_1","NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4, FP8 (KV cache)","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, Dynamo)","GB200-NVL72_GB200-186GB_aarch64x4_Dynamo","qwen3-vl-235b-a22b","VLM","Offline","58.395875238853066","0.7870702100264627","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x4_Dynamo/qwen3-vl-235b-a22b/Offline/performance/run_1","NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","NVFP4, FP8 (KV cache)","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, Dynamo)","GB200-NVL72_GB200-186GB_aarch64x4_Dynamo","qwen3-vl-235b-a22b","VLM","Server","49.6962883240102","0.7878435998292614","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x4_Dynamo/qwen3-vl-235b-a22b/Server/performance/run_1","NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4, FP8 (KV cache)","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x4_TRT","gpt-oss-120b","gpt-oss-120b","Offline","50912.9","exact_match: 83.691","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x4_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Crusoe","available","closed","datacenter","NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x4_TRT","gpt-oss-120b","gpt-oss-120b","Server","50613.5","exact_match: 83.642","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/Crusoe/results/GB200-NVL72_GB200-186GB_aarch64x4_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","PowerEdge R770AP_128C 24x96GB (6400 MT/s)","1-node-2S-GNR_128C","rgat","rgat","Offline","21835.4","acc: 72.578","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Dell/results/1-node-2S-GNR_128C/rgat/Offline/performance/run_1","PyTorch","Ubuntu 22.04.5 LTS","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Dell","available","closed","datacenter","PowerEdge R770AP_128C 24x96GB (6400 MT/s)","1-node-2S-GNR_128C","whisper","whisper","Offline","2275.93","ACCURACY: 97.89143156192017","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Dell/results/1-node-2S-GNR_128C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 22.04.5 LTS","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Dell","available","closed","datacenter","PowerEdge R770_86C 16x64GB (6400 MT/s)","1-node-2S-GNR_86C","llama3_1-8b","llama3.1-8b","Offline","1227.99","ROUGE1: 38.6229  ROUGE2: 15.8107  ROUGEL: 24.4082  ROUGELSUM: 35.6517  GEN_LEN: 8188478","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Dell/results/1-node-2S-GNR_86C/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04 LTS","Dell PowerEdge R770. vLLM, FP8 KV Cache","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Dell","available","closed","datacenter","PowerEdge R770_86C 16x64GB (6400 MT/s)","1-node-2S-GNR_86C","llama3_1-8b","llama3.1-8b","Server","516.012","ROUGE1: 38.6453  ROUGE2: 15.8841  ROUGEL: 24.4777  ROUGELSUM: 35.7016  GEN_LEN: 8176464","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Dell/results/1-node-2S-GNR_86C/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04 LTS","Dell PowerEdge R770. vLLM, FP8 KV Cache","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Dell","available","closed","datacenter","PowerEdge R770_86C 16x64GB (6400 MT/s)","1-node-2S-GNR_86C","rgat","rgat","Offline","16042.8","acc: 72.570","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Dell/results/1-node-2S-GNR_86C/rgat/Offline/performance/run_1","PyTorch","Ubuntu 24.04 LTS","Dell PowerEdge R770. vLLM, FP8 KV Cache","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Dell","available","closed","datacenter","PowerEdge R770_86C 16x64GB (6400 MT/s)","1-node-2S-GNR_86C","whisper","whisper","Offline","1539.49","ACCURACY: 97.89904372234645","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Dell/results/1-node-2S-GNR_86C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04 LTS","Dell PowerEdge R770. vLLM, FP8 KV Cache","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Dell","available","closed","edge","Dell Pro Max (1x GB10)","Dell_Pro_Max_GB10x1","yolo-95","yolo-95","multistream","105.162072","mAP: 53.113","1","Cortex-X925","20","1","NVIDIA GB10","1","1","./closed/Dell/results/Dell_Pro_Max_GB10x1/yolo-95/multistream/performance/run_1","CUDA 13.0","Ubuntu 24.04.3 LTS","Dell Pro Max with GB10","closed","0","v6.1","1","False","Latency (ms)","fp4","None",""
"Dell","available","closed","edge","Dell Pro Max (1x GB10)","Dell_Pro_Max_GB10x1","yolo-95","yolo-95","offline","78.7257","mAP: 53.113","1","Cortex-X925","20","1","NVIDIA GB10","1","1","./closed/Dell/results/Dell_Pro_Max_GB10x1/yolo-95/offline/performance/run_1","CUDA 13.0","Ubuntu 24.04.3 LTS","Dell Pro Max with GB10","closed","0","v6.1","1","False","Samples/s","fp4","None",""
"Dell","available","closed","edge","Dell Pro Max (1x GB10)","Dell_Pro_Max_GB10x1","yolo-95","yolo-95","singlestream","13.865536","mAP: 53.113","1","Cortex-X925","20","1","NVIDIA GB10","1","1","./closed/Dell/results/Dell_Pro_Max_GB10x1/yolo-95/singlestream/performance/run_1","CUDA 13.0","Ubuntu 24.04.3 LTS","Dell Pro Max with GB10","closed","0","v6.1","1","False","Latency (ms)","fp4","None",""
"Dell","available","closed","edge","Dell Pro Max (1x GB10)","Dell_Pro_Max_GB10x1","yolo-99","yolo-99","multistream","105.162072","mAP: 53.113","1","Cortex-X925","20","1","NVIDIA GB10","1","1","./closed/Dell/results/Dell_Pro_Max_GB10x1/yolo-99/multistream/performance/run_1","CUDA 13.0","Ubuntu 24.04.3 LTS","Dell Pro Max with GB10","closed","0","v6.1","1","False","Latency (ms)","fp4","None",""
"Dell","available","closed","edge","Dell Pro Max (1x GB10)","Dell_Pro_Max_GB10x1","yolo-99","yolo-99","offline","78.7257","mAP: 53.113","1","Cortex-X925","20","1","NVIDIA GB10","1","1","./closed/Dell/results/Dell_Pro_Max_GB10x1/yolo-99/offline/performance/run_1","CUDA 13.0","Ubuntu 24.04.3 LTS","Dell Pro Max with GB10","closed","0","v6.1","1","False","Samples/s","fp4","None",""
"Dell","available","closed","edge","Dell Pro Max (1x GB10)","Dell_Pro_Max_GB10x1","yolo-99","yolo-99","singlestream","13.865536","mAP: 53.113","1","Cortex-X925","20","1","NVIDIA GB10","1","1","./closed/Dell/results/Dell_Pro_Max_GB10x1/yolo-99/singlestream/performance/run_1","CUDA 13.0","Ubuntu 24.04.3 LTS","Dell Pro Max with GB10","closed","0","v6.1","1","False","Latency (ms)","fp4","None",""
"Dell","available","closed","datacenter","PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","XE7740_RTXPro6000_PCIe_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","29796.7","ROUGE1: 44.7408  ROUGE2: 22.3357  ROUGEL: 29.1354  TOKENS_PER_SAMPLE: 274.4","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Dell/results/XE7740_RTXPro6000_PCIe_96GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","XE7740_RTXPro6000_PCIe_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","28973.3","ROUGE1: 44.6929  ROUGE2: 22.2906  ROUGEL: 29.0805  TOKENS_PER_SAMPLE: 267.3","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Dell/results/XE7740_RTXPro6000_PCIe_96GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA RTX PRO 6000 Blackwell Server Edition TGP 600W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","PowerEdge XE7745 (8x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","XE7745_RTXPRO4500_32GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","4374.06","ROUGE1: 44.7919  ROUGE2: 22.3659  ROUGEL: 29.1627  TOKENS_PER_SAMPLE: 273.2","1","AMD EPYC 9655","1","96","NVIDIA RTX PRO 4500 Blackwell","8","8","./closed/Dell/results/XE7745_RTXPRO4500_32GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.3","Ubuntu 24.04.3","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","PowerEdge XE7745 (8x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","XE7745_RTXPRO4500_32GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","3012.33","ROUGE1: 44.7861  ROUGE2: 22.3547  ROUGEL: 29.1687  TOKENS_PER_SAMPLE: 271.3","1","AMD EPYC 9655","1","96","NVIDIA RTX PRO 4500 Blackwell","8","8","./closed/Dell/results/XE7745_RTXPRO4500_32GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.3","Ubuntu 24.04.3","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","Dell PowerEdge XE9712","XE9712x18","wan-2.2-t2v-a14b","text_to_video","Offline","0.6455254875195523","0.7009552217488585","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Dell/results/XE9712x18/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"Dell","available","closed","datacenter","Dell PowerEdge XE9712","XE9712x18","wan-2.2-t2v-a14b","text_to_video","SingleStream","10.585000540652779","0.6970976320484682","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Dell/results/XE9712x18/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"Dell","available","closed","datacenter","Dell XE9780 (8x B300-SXM-270GB, TensorRT)","XE9780-B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","109257.0","exact_match: 83.610","1","Intel(R) Xeon(R) 6747P","2","96","Nvidia B300-SXM-270GB","8","8","./closed/Dell/results/XE9780-B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, Dell Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","Dell XE9780 (8x B300-SXM-270GB, TensorRT)","XE9780-B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","103332.0","exact_match: 83.460","1","Intel(R) Xeon(R) 6747P","2","96","Nvidia B300-SXM-270GB","8","8","./closed/Dell/results/XE9780-B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, Dell Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","Dell XE9780 (8x B300-SXM-270GB, TensorRT)","XE9780-B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","104006.0","ROUGE1: 44.9129  ROUGE2: 22.4979  ROUGEL: 29.3269  TOKENS_PER_SAMPLE: 271.0","1","Intel(R) Xeon(R) 6747P","2","96","Nvidia B300-SXM-270GB","8","8","./closed/Dell/results/XE9780-B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, Dell Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","Dell XE9780 (8x B300-SXM-270GB, TensorRT)","XE9780-B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","96180.6","ROUGE1: 44.8204  ROUGE2: 22.399  ROUGEL: 29.2127  TOKENS_PER_SAMPLE: 272.0","1","Intel(R) Xeon(R) 6747P","2","96","Nvidia B300-SXM-270GB","8","8","./closed/Dell/results/XE9780-B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, Dell Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","gpt-oss-120b","gpt-oss-120b","Offline","117804.0","exact_match: 84.217","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","gpt-oss-120b","gpt-oss-120b","Server","110188.0","exact_match: 83.843","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","llama2-70b-99.9","llama2-70b-99.9","Interactive","73238.1","ROUGE1: 44.4905  ROUGE2: 22.144  ROUGEL: 28.908  TOKENS_PER_SAMPLE: 284.4","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/llama2-70b-99.9/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","llama2-70b-99.9","llama2-70b-99.9","Offline","103792.0","ROUGE1: 44.5322  ROUGE2: 22.1855  ROUGEL: 28.9228  TOKENS_PER_SAMPLE: 284.4","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/llama2-70b-99.9/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","llama2-70b-99.9","llama2-70b-99.9","Server","103932.0","ROUGE1: 44.5341  ROUGE2: 22.1716  ROUGEL: 28.9333  TOKENS_PER_SAMPLE: 284.0","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/llama2-70b-99.9/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","llama3.1-8b","llama3.1-8b","Interactive","130540.0","ROUGE1: 38.561  ROUGE2: 16.0262  ROUGEL: 24.4382  ROUGELSUM: 35.6995  GEN_LEN: 8150427","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/llama3.1-8b/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","llama3.1-8b","llama3.1-8b","Offline","158458.0","ROUGE1: 38.5617  ROUGE2: 16.0263  ROUGEL: 24.4381  ROUGELSUM: 35.6994  GEN_LEN: 8151245","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/llama3.1-8b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell","available","closed","datacenter","XE9785L_8xMI355X","XE9785L_8xMI355X","llama3.1-8b","llama3.1-8b","Server","148563.0","ROUGE1: 38.5211  ROUGE2: 16.0095  ROUGEL: 24.4208  ROUGELSUM: 35.662  GEN_LEN: 8168331","1","AMD EPYC 9655","2","96","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell/results/XE9785L_8xMI355X/llama3.1-8b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell","available","closed","datacenter","Dell XE9785L (8x B300-SXM-270GB, TensorRT)","XE9785L_x86_B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","113101.0","ROUGE1: 44.8017  ROUGE2: 22.3823  ROUGEL: 29.1656  TOKENS_PER_SAMPLE: 272.7","1","AMD EPYC 9965","2","192","NVIDIA B300-SXM-270GB","8","8","./closed/Dell/results/XE9785L_x86_B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell","available","closed","datacenter","Dell XE9785L (8x B300-SXM-270GB, TensorRT)","XE9785L_x86_B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106898.0","ROUGE1: 44.8186  ROUGE2: 22.3916  ROUGEL: 29.1862  TOKENS_PER_SAMPLE: 272.3","1","AMD EPYC 9965","2","192","NVIDIA B300-SXM-270GB","8","8","./closed/Dell/results/XE9785L_x86_B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","dlrm-v3","dlrm-v3","Offline","4837.63","DLRM_NE: 0.8677279311862482  DLRM_ACC: 0.6964196251791978  DLRM_AUC: 0.7862870079033047","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/dlrm-v3/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Samples/s","int8,bf16","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","dlrm-v3","dlrm-v3","Server","4503.53","DLRM_NE: 0.8677277210685596  DLRM_ACC: 0.6964198708391458  DLRM_AUC: 0.7862873761281074","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/dlrm-v3/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Queries/s","int8,bf16","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","gpt-oss-120b","gpt-oss-120b","Offline","48971.1","exact_match: 83.287","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","gpt-oss-120b","gpt-oss-120b","Server","40867.9","exact_match: 83.631","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","llama2-70b-99.9","llama2-70b-99.9","Interactive","22860.5","ROUGE1: 44.4031  ROUGE2: 22.0641  ROUGEL: 28.8533  TOKENS_PER_SAMPLE: 288.3","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/llama2-70b-99.9/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","llama2-70b-99.9","llama2-70b-99.9","Offline","42286.2","ROUGE1: 44.5012  ROUGE2: 22.1742  ROUGEL: 28.9465  TOKENS_PER_SAMPLE: 284.2","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/llama2-70b-99.9/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","llama2-70b-99.9","llama2-70b-99.9","Server","41830.0","ROUGE1: 44.4477  ROUGE2: 22.0889  ROUGEL: 28.8674  TOKENS_PER_SAMPLE: 287.4","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/llama2-70b-99.9/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","llama3.1-8b","llama3.1-8b","Interactive","62723.7","ROUGE1: 38.5629  ROUGE2: 16.0233  ROUGEL: 24.4351  ROUGELSUM: 35.6984  GEN_LEN: 8150299","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/llama3.1-8b/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","llama3.1-8b","llama3.1-8b","Offline","73643.0","ROUGE1: 38.5604  ROUGE2: 16.0125  ROUGEL: 24.4415  ROUGELSUM: 35.6907  GEN_LEN: 8149596","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/llama3.1-8b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","llama3.1-8b","llama3.1-8b","Server","72305.1","ROUGE1: 38.5633  ROUGE2: 16.0282  ROUGEL: 24.4369  ROUGELSUM: 35.7015  GEN_LEN: 8151565","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/llama3.1-8b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","wan-2.2-t2v-a14b","text_to_video","Offline","0.0345875","vbench_score: 69.9361","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/wan-2.2-t2v-a14b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Samples/s","bf16, fp8, fp4","None",""
"Dell_AMD","available","closed","datacenter","XE7745_8xMI350P_2xEPYC_9455","XE7745_8xMI350P_2xEPYC-9455","wan-2.2-t2v-a14b","text_to_video","SingleStream","41.837923108000005","vbench_score: 70.3648","1","AMD EPYC 9455","2","48","AMD Instinct MI350P","8","8","./closed/Dell_AMD/results/XE7745_8xMI350P_2xEPYC-9455/wan-2.2-t2v-a14b/SingleStream/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 24.04.4 LTS (kernel 6.8.0-136-generic)","8x AMD Instinct MI350P (gfx950), 144GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Latency (s)","bf16, fp8, fp4","None",""
"Dell_MangoBoost","available","closed","datacenter","Heterogeneous Four-Cluster (2x8x MI300X + 2x8x MI355X, LLMBoost)","2x8xMI355X_2x8xMI300X","gpt-oss-120b","gpt-oss-120b","Offline","285454.0","exact_match: 84.250","4","AMD EPYC 9534 (on MI300X node) and AMD EPYC 9965 (on MI355X node)","2","64(AMD EPYC 9534) and 192(AMD EPYC 9965)","AMD Instinct MI300X 192GB HBM3 (x2x8) and AMD Instinct MI355X 288GB HBM3e (x2x8)","8","32","./closed/Dell_MangoBoost/results/2x8xMI355X_2x8xMI300X/gpt-oss-120b/Offline/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Heterogeneous Four-Cluster with LLMBoost, vLLM 0.22.1 on MI355X, and vLLM 0.24.0 on MI300X","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","Heterogeneous Four-Cluster (2x8x MI300X + 2x8x MI355X, LLMBoost)","2x8xMI355X_2x8xMI300X","gpt-oss-120b","gpt-oss-120b","Server","253501.0","exact_match: 83.574","4","AMD EPYC 9534 (on MI300X node) and AMD EPYC 9965 (on MI355X node)","2","64(AMD EPYC 9534) and 192(AMD EPYC 9965)","AMD Instinct MI300X 192GB HBM3 (x2x8) and AMD Instinct MI355X 288GB HBM3e (x2x8)","8","32","./closed/Dell_MangoBoost/results/2x8xMI355X_2x8xMI300X/gpt-oss-120b/Server/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Heterogeneous Four-Cluster with LLMBoost, vLLM 0.22.1 on MI355X, and vLLM 0.24.0 on MI300X","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","PowerEdge XE9785","XE9785","gpt-oss-120b","gpt-oss-120b","Interactive","28265.9","exact_match: 83.676","1","AMD EPYC 9965","2","192","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell_MangoBoost/results/XE9785/gpt-oss-120b/Interactive/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Intra-Node P/D Disaggregation for Interactive Scenario, and vLLM 0.22.1","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","PowerEdge XE9785","XE9785","gpt-oss-120b","gpt-oss-120b","Offline","115294.0","exact_match: 83.720","1","AMD EPYC 9965","2","192","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell_MangoBoost/results/XE9785/gpt-oss-120b/Offline/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Intra-Node P/D Disaggregation for Interactive Scenario, and vLLM 0.22.1","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","PowerEdge XE9785","XE9785","gpt-oss-120b","gpt-oss-120b","Server","109065.0","exact_match: 83.806","1","AMD EPYC 9965","2","192","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell_MangoBoost/results/XE9785/gpt-oss-120b/Server/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Intra-Node P/D Disaggregation for Interactive Scenario, and vLLM 0.22.1","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","PowerEdge XE9785L","XE9785L","gpt-oss-120b","gpt-oss-120b","Interactive","29585.5","exact_match: 84.127","1","AMD EPYC 9965","2","192","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell_MangoBoost/results/XE9785L/gpt-oss-120b/Interactive/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Intra-Node P/D Disaggregation for Interactive Scenario, and vLLM 0.22.1","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","PowerEdge XE9785L","XE9785L","gpt-oss-120b","gpt-oss-120b","Offline","116953.0","exact_match: 83.643","1","AMD EPYC 9965","2","192","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell_MangoBoost/results/XE9785L/gpt-oss-120b/Offline/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Intra-Node P/D Disaggregation for Interactive Scenario, and vLLM 0.22.1","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Dell_MangoBoost","available","closed","datacenter","PowerEdge XE9785L","XE9785L","gpt-oss-120b","gpt-oss-120b","Server","110153.0","exact_match: 83.598","1","AMD EPYC 9965","2","192","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Dell_MangoBoost/results/XE9785L/gpt-oss-120b/Server/performance/run_1","ROCm","Ubuntu 22.04.5 LTS (jammy)","Intra-Node P/D Disaggregation for Interactive Scenario, and vLLM 0.22.1","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Fujitsu","available","closed","datacenter","PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8, TensorRT)","PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","26626.5","ROUGE1: 44.7182  ROUGE2: 22.3154  ROUGEL: 29.104  TOKENS_PER_SAMPLE: 273.6","1","Intel(R) Xeon(R) Platinum 8592+","2","64","NVIDIA RTX PRO 6000 Blackwell","8","8","./closed/Fujitsu/results/PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.16, CUDA 13.2, cuDNN 9.21","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Fujitsu","available","closed","datacenter","PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8, TensorRT)","PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","27340.9","ROUGE1: 44.7475  ROUGE2: 22.3697  ROUGEL: 29.1476  TOKENS_PER_SAMPLE: 274.5","1","Intel(R) Xeon(R) Platinum 8592+","2","64","NVIDIA RTX PRO 6000 Blackwell","8","8","./closed/Fujitsu/results/PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.16, CUDA 13.2, cuDNN 9.21","Ubuntu 24.04","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","72445.1","exact_match: 81.24430264357338  TOKENS_PER_SAMPLE: 3793.010711030082","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","70495.7","exact_match: 81.26709206927985  TOKENS_PER_SAMPLE: 3788.4774384685506","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","111625.0","exact_match: 83.316","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","111218.0","exact_match: 83.071","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","115034.0","ROUGE1: 44.803  ROUGE2: 22.3802  ROUGEL: 29.1563  TOKENS_PER_SAMPLE: 272.6","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","114832.0","ROUGE1: 44.8381  ROUGE2: 22.394  ROUGEL: 29.1852  TOKENS_PER_SAMPLE: 272.5","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Interactive","12.485432132404888","0.7879417016152472","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","nvfp4, fp8-kv","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","131.2852847183584","0.7870700970178708","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","nvfp4, fp8-kv","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Server","111.7532977914224","0.7880503014737441","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","nvfp4, fp8-kv","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.06291526723554443","0.7028431480110648","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"GigaComputing","available","closed","datacenter","G894-AD3-AAX7","G894-AD3-AAX7_B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","19.079404388444445","0.7036278864900404","1","Intel(R) Xeon(R) 6979P","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-AD3-AAX7_B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","72327.1","exact_match: 80.9936189608022  TOKENS_PER_SAMPLE: 3803.4314038286234","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","70771.7","exact_match: 80.76572470373746  TOKENS_PER_SAMPLE: 3851.8625797629898","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","111541.0","exact_match: 82.928","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","110592.0","exact_match: 83.965","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","114826.0","ROUGE1: 44.8002  ROUGE2: 22.3763  ROUGEL: 29.1593  TOKENS_PER_SAMPLE: 272.5","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","114578.0","ROUGE1: 44.8817  ROUGE2: 22.4296  ROUGEL: 29.2316  TOKENS_PER_SAMPLE: 271.9","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Interactive","12.881779740805642","0.787262910148138","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","NVFP4 weights with FP8 KV cache","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","132.0547880616454","0.7875530423168103","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","NVFP4 weights with FP8 KV cache","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Server","115.78393993840092","0.7880234463575065","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","NVFP4 weights with FP8 KV cache","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.0625123079679506","0.7048583781540092","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"GigaComputing","available","closed","datacenter","G894-ZD3-AAX7","G894-ZD3-AAX7_B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","18.051798671840277","0.6987459561814395","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/GigaComputing/results/G894-ZD3-AAX7_B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 22.04.4","B300 TGP 1100W","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"Google","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Interactive","232160.0","exact_match: 80.9024612579763  TOKENS_PER_SAMPLE: 3809.1959890610756","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/Google/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Interactive/performance/run_1","PyTorch 2.10.0a0+b4e4ee81d3.nv25.12, TensorRT 10.14.1.48, CUDA Toolkit 13.1, cuDNN 9.17.0, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU. DeepSeek-R1 used the pinned aarch64 MLPerf TensorRT-LLM container. Runtime logs show MNNVL/NVLinkOneSided selection for Offline, Server, and Interactive; colocated Offline/Server also initialized the NIXL UCX backend, while disaggregated Interactive used a UCX KV-cache transceiver between context and generation services.","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Google","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","510781.0","exact_match: 81.19872379216044  TOKENS_PER_SAMPLE: 3852.355970829535","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/Google/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","PyTorch 2.10.0a0+b4e4ee81d3.nv25.12, TensorRT 10.14.1.48, CUDA Toolkit 13.1, cuDNN 9.17.0, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU. DeepSeek-R1 used the pinned aarch64 MLPerf TensorRT-LLM container. Runtime logs show MNNVL/NVLinkOneSided selection for Offline, Server, and Interactive; colocated Offline/Server also initialized the NIXL UCX backend, while disaggregated Interactive used a UCX KV-cache transceiver between context and generation services.","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Google","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","388590.0","exact_match: 81.01640838650866  TOKENS_PER_SAMPLE: 3795.0934366453967","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/Google/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","PyTorch 2.10.0a0+b4e4ee81d3.nv25.12, TensorRT 10.14.1.48, CUDA Toolkit 13.1, cuDNN 9.17.0, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU. DeepSeek-R1 used the pinned aarch64 MLPerf TensorRT-LLM container. Runtime logs show MNNVL/NVLinkOneSided selection for Offline, Server, and Interactive; colocated Offline/Server also initialized the NIXL UCX backend, while disaggregated Interactive used a UCX KV-cache transceiver between context and generation services.","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","NVIDIA GB200 NVL72 by HPE (8x GB200-186GB_aarch64, TensorRT)","HPE_GB200-NVL72_GB200-186GB_aarch64x8","gpt-oss-120b","gpt-oss-120b","Offline","101453.0","exact_match: 83.727","2","NVIDIA Grace CPU (Arm Neoverse-V2)","2","144","NVIDIA GB200","4","8","./closed/HPE/results/HPE_GB200-NVL72_GB200-186GB_aarch64x8/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k","NVIDIA GB200 NVL72 platform by HPE; 1200 W GPU power limit. Two dep8 replicas across two nodes; CentML DeepSeek-R1 NVFP4 v2 checkpoint","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","NVIDIA GB200 NVL72 by HPE (8x GB200-186GB_aarch64, TensorRT)","HPE_GB200-NVL72_GB200-186GB_aarch64x8","gpt-oss-120b","gpt-oss-120b","Server","100714.0","exact_match: 83.485","2","NVIDIA Grace CPU (Arm Neoverse-V2)","2","144","NVIDIA GB200","4","8","./closed/HPE/results/HPE_GB200-NVL72_GB200-186GB_aarch64x8/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k","NVIDIA GB200 NVL72 platform by HPE; 1200 W GPU power limit. Two dep8 replicas across two nodes; CentML DeepSeek-R1 NVFP4 v2 checkpoint","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","edge","HPE ProLiant DL145 Gen11 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT)","HPE_PROLIANT_DL145_RTX_PRO_4500_PCIE_32GBx1","qwen3.6-27b","agentic edge","SingleStream","2757.954404037736","0.8613065326633166","1","AMD EPYC 8535P","1","64","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL145_RTX_PRO_4500_PCIE_32GBx1/qwen3.6-27b/SingleStream/performance/run_1","llama.cpp","Ubuntu 24.04.3","","closed","0","v6.1","1","False","Latency (ms)","int4","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL345 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT)","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx1","yolo-95","yolo-95","multistream","676.268814","mAP: 53.114","1","AMD EPYC 9375F","1","32","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx1/yolo-95/multistream/performance/run_1","PyTorch v2.13.0","Ubuntu 24.04.4","","closed","0","v6.1","1","False","Latency (ms)","fp32","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL345 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT)","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx1","yolo-95","yolo-95","offline","146.207","mAP: 53.109","1","AMD EPYC 9375F","1","32","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx1/yolo-95/offline/performance/run_1","PyTorch v2.13.0","Ubuntu 24.04.4","","closed","0","v6.1","1","False","Samples/s","fp32","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL345 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT)","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx1","yolo-95","yolo-95","singlestream","76.664039","mAP: 53.114","1","AMD EPYC 9375F","1","32","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx1/yolo-95/singlestream/performance/run_1","PyTorch v2.13.0","Ubuntu 24.04.4","","closed","0","v6.1","1","False","Latency (ms)","fp32","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL345 Gen12 (4x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_AGENTIC","qwen3.6-27b","agentic edge","SingleStream","2616.1590054885796","0.864321608040201","1","AMD EPYC 9375F","1","32","NVIDIA RTX PRO 4500 Blackwell Server Edition","4","4","./closed/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_AGENTIC/qwen3.6-27b/SingleStream/performance/run_1","llama.cpp","Ubuntu 24.04.4","","closed","0","v6.1","1","False","Latency (ms)","int4","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL345 Gen12 (4x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_TRT","whisper","whisper","Offline","3685.7","ACCURACY: 97.84824462061155","1","AMD EPYC 9375F","1","32","NVIDIA RTX PRO 4500 Blackwell Server Edition","4","4","./closed/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_TRT/whisper/Offline/performance/run_1","TensorRT 10.15.1.29, CUDA 13.2","Ubuntu 24.04.4","","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","gpt-oss-120b","gpt-oss-120b","Offline","15256.0","exact_match: 82.714","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","gpt-oss-120b","gpt-oss-120b","Server","14581.7","exact_match: 82.341","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/gpt-oss-120b/Server/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","llama2-70b-99.9","llama2-70b-99.9","Offline","29284.8","ROUGE1: 44.7216  ROUGE2: 22.3081  ROUGEL: 29.0908  TOKENS_PER_SAMPLE: 274.8","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","llama2-70b-99.9","llama2-70b-99.9","Server","28643.3","ROUGE1: 44.7706  ROUGE2: 22.352  ROUGEL: 29.1377  TOKENS_PER_SAMPLE: 272.0","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","llama3.1-8b","llama3.1-8b","Interactive","39986.3","ROUGE1: 38.7645  ROUGE2: 16.0515  ROUGEL: 24.5586  ROUGELSUM: 35.6824  GEN_LEN: 8173226","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/llama3.1-8b/Interactive/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","llama3.1-8b","llama3.1-8b","Offline","50965.3","ROUGE1: 38.7974  ROUGE2: 16.0868  ROUGEL: 24.5879  ROUGELSUM: 35.7199  GEN_LEN: 8172782","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/llama3.1-8b/Offline/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","llama3.1-8b","llama3.1-8b","Server","49523.3","ROUGE1: 38.7821  ROUGE2: 16.0772  ROUGEL: 24.5857  ROUGELSUM: 35.7074  GEN_LEN: 8172428","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/llama3.1-8b/Server/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8","whisper","whisper","Offline","19278.3","ACCURACY: 97.88345689351618","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/HPE/results/HPE_PROLIANT_DL380A_RTX_PRO_6000_PCIE_96GBx8/whisper/Offline/performance/run_1","TensorRT 10.16.1.11, CUDA 13.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Samples/s","fp4","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL380 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB)","HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT","yolo-95","yolo-95","multistream","116.101325","mAP: 53.112","1","Intel(R) Xeon(R) 6762P","2","64","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT/yolo-95/multistream/performance/run_1","PyTorch v2.13.0","Ubuntu 24.04.3","","closed","0","v6.1","1","False","Latency (ms)","fp32","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL380 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB)","HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT","yolo-95","yolo-95","offline","72.1855","mAP: 53.112","1","Intel(R) Xeon(R) 6762P","2","64","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT/yolo-95/offline/performance/run_1","PyTorch v2.13.0","Ubuntu 24.04.3","","closed","0","v6.1","1","False","Samples/s","fp32","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL380 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB)","HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT","yolo-95","yolo-95","singlestream","14.136709","mAP: 53.112","1","Intel(R) Xeon(R) 6762P","2","64","NVIDIA RTX PRO 4500 Blackwell Server Edition","1","1","./closed/HPE/results/HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT/yolo-95/singlestream/performance/run_1","PyTorch v2.13.0","Ubuntu 24.04.3","","closed","0","v6.1","1","False","Latency (ms)","fp32","None",""
"HPE","available","closed","edge","HPE ProLiant Compute DL380 Gen12 (5x NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB, TensorRT)","HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx5_TRT","whisper","whisper","Offline","4719.46","ACCURACY: 97.85585808502336","1","Intel(R) Xeon(R) 6762P","2","64","NVIDIA RTX PRO 4500 Blackwell Server Edition","5","5","./closed/HPE/results/HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx5_TRT/whisper/Offline/performance/run_1","TensorRT 10.11.0.33, CUDA 13.2","Ubuntu 24.04.3","","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","gpt-oss-120b","gpt-oss-120b","Offline","117355.0","exact_match: 83.687","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","gpt-oss-120b","gpt-oss-120b","Server","108836.0","exact_match: 83.864","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","llama2-70b-99.9","llama2-70b-99.9","Interactive","72977.2","ROUGE1: 44.5152  ROUGE2: 22.1564  ROUGEL: 28.9213  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/llama2-70b-99.9/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","llama2-70b-99.9","llama2-70b-99.9","Offline","102991.0","ROUGE1: 44.5406  ROUGE2: 22.1922  ROUGEL: 28.9322  TOKENS_PER_SAMPLE: 284.2","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/llama2-70b-99.9/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","llama2-70b-99.9","llama2-70b-99.9","Server","101979.0","ROUGE1: 44.5325  ROUGE2: 22.1714  ROUGEL: 28.9308  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/llama2-70b-99.9/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","llama3.1-8b","llama3.1-8b","Interactive","130613.0","ROUGE1: 38.567  ROUGE2: 16.0282  ROUGEL: 24.4449  ROUGELSUM: 35.7031  GEN_LEN: 8150301","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/llama3.1-8b/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","llama3.1-8b","llama3.1-8b","Offline","157694.0","ROUGE1: 38.5626  ROUGE2: 16.0276  ROUGEL: 24.4431  ROUGELSUM: 35.699  GEN_LEN: 8151086","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/llama3.1-8b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","llama3.1-8b","llama3.1-8b","Server","148118.0","ROUGE1: 38.5316  ROUGE2: 16.0159  ROUGEL: 24.4259  ROUGELSUM: 35.6716  GEN_LEN: 8162750","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/llama3.1-8b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","wan-2.2-t2v-a14b","text_to_video","Offline","0.0707035","vbench_score: 70.1048","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/wan-2.2-t2v-a14b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Samples/s","bf16, fp8, fp4","None",""
"HPE","available","closed","datacenter","HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","HPE_XD685_MI355X_288GBx8_vLLM","wan-2.2-t2v-a14b","text_to_video","SingleStream","16.397060151","vbench_score: 70.3025","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/HPE/results/HPE_XD685_MI355X_288GBx8_vLLM/wan-2.2-t2v-a14b/SingleStream/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.5 LTS (kernel 5.15.0-186-generic)","HPE ProLiant Compute XD685; 8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Latency (s)","bf16, fp8, fp4","None",""
"HPE","available","closed","datacenter","HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","HPE_XD690_B300_SXM_270GBx16_TRT","deepseek-r1","deepseek-r1","Offline","136854.0","exact_match: 81.24430264357338  TOKENS_PER_SAMPLE: 3826.166362807657","2","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","16","./closed/HPE/results/HPE_XD690_B300_SXM_270GBx16_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.2","RHEL 9.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","HPE_XD690_B300_SXM_270GBx16_TRT","deepseek-r1","deepseek-r1","Server","80536.9","exact_match: 80.92525068368278  TOKENS_PER_SAMPLE: 3807.0186873290795","2","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","16","./closed/HPE/results/HPE_XD690_B300_SXM_270GBx16_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.2","RHEL 9.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","HPE_XD690_B300_SXM_270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","114569.0","exact_match: 83.235","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/HPE/results/HPE_XD690_B300_SXM_270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.2","RHEL 9.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","closed","datacenter","HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","HPE_XD690_B300_SXM_270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","113685.0","exact_match: 83.448","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/HPE/results/HPE_XD690_B300_SXM_270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.2","RHEL 9.4","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Intel","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","llama3_1-8b","llama3.1-8b","Offline","1916.74","ROUGE1: 38.6418  ROUGE2: 15.8693  ROUGEL: 24.4809  ROUGELSUM: 35.6703  GEN_LEN: 8194131","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Intel/results/1-node-2S-GNR_128C/llama3_1-8b/Offline/performance/run_1","PyTorch","CentOS Stream 9","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","llama3_1-8b","llama3.1-8b","Server","1139.51","ROUGE1: 38.6681  ROUGE2: 15.9242  ROUGEL: 24.5144  ROUGELSUM: 35.7046  GEN_LEN: 8206137","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Intel/results/1-node-2S-GNR_128C/llama3_1-8b/Server/performance/run_1","PyTorch","CentOS Stream 9","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","rgat","rgat","Offline","24573.7","acc: 72.541","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Intel/results/1-node-2S-GNR_128C/rgat/Offline/performance/run_1","PyTorch","CentOS Stream 9","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","whisper","whisper","Offline","2319.02","ACCURACY: 97.89143156192017","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Intel/results/1-node-2S-GNR_128C/whisper/Offline/performance/run_1","PyTorch","CentOS Stream 9","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-2S-Xeon_6787P","1-node-2S-Xeon_6787P","e2e-rag-db","end to end vector-database","Offline","22.6141","E2E_ACCURACY: 98.1952","1","Intel(R) Xeon(R) 6787P","2","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-2S-Xeon_6787P/e2e-rag-db/Offline/performance/run_1","vLLM 0.22.1rc1.dev189+ga55fccfc7.d20260604, PyTorch 2.11.0+cpu, FAISS 1.14.3","Ubuntu 24.04.4 LTS (kernel 7.0.0-28-generic)","e2e-rag-qna: gpt-oss-120b on 4x Intel Arc Pro B70 (TP=4); gpt-oss-20b grader, e5-base-v2 embedding and colbertv2.0 reranking on CPU. e2e-rag-db: CPU-only vector-database build, no accelerator used.. e2e-rag-qna: async multi-shot RAG pipeline with pinned CPU embedding/FAISS and reranking services. e2e-rag-db: vector DB built from the frozen docs.tar.gz Wikipedia corpus (2515 HTML pages).","closed","0","v6.1","0","False","Samples/s","fp32","None",""
"Intel","available","closed","datacenter","1-node-2S-Xeon_6787P","1-node-2S-Xeon_6787P","e2e-rag-qna","end to end question-answering","Offline","0.464076","E2E_ACCURACY: 35.9223","1","Intel(R) Xeon(R) 6787P","2","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-2S-Xeon_6787P/e2e-rag-qna/Offline/performance/run_1","vLLM 0.22.1rc1.dev189+ga55fccfc7.d20260604, PyTorch 2.11.0+cpu, FAISS 1.14.3","Ubuntu 24.04.4 LTS (kernel 7.0.0-28-generic)","e2e-rag-qna: gpt-oss-120b on 4x Intel Arc Pro B70 (TP=4); gpt-oss-20b grader, e5-base-v2 embedding and colbertv2.0 reranking on CPU. e2e-rag-db: CPU-only vector-database build, no accelerator used.. e2e-rag-qna: async multi-shot RAG pipeline with pinned CPU embedding/FAISS and reranking services. e2e-rag-db: vector DB built from the frozen docs.tar.gz Wikipedia corpus (2515 HTML pages).","closed","0","v6.1","0","False","Tasks/s","mxfp4 (LLMs), fp32 (embedding/reranker)","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","gpt-oss-120b","gpt-oss-120b","Offline","1956.4","exact_match: 83.674","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/gpt-oss-120b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","MXFP4","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","gpt-oss-120b","gpt-oss-120b","Server","1296.87","exact_match: 83.489","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/gpt-oss-120b/Server/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","MXFP4","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama2-70b-99.9","llama2-70b-99.9","Offline","2464.98","ROUGE1: 44.5947  ROUGE2: 22.2074  ROUGEL: 28.8264  TOKENS_PER_SAMPLE: 290.8","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/llama2-70b-99.9/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama2-70b-99.9","llama2-70b-99.9","Server","1745.48","ROUGE1: 44.5937  ROUGE2: 22.2052  ROUGEL: 28.8276  TOKENS_PER_SAMPLE: 290.8","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/llama2-70b-99.9/Server/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama3_1-8b","llama3.1-8b","Offline","4613.28","ROUGE1: 38.843  ROUGE2: 16.0878  ROUGEL: 24.6429  ROUGELSUM: 35.9449  GEN_LEN: 8127480","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama3_1-8b","llama3.1-8b","Server","4173.25","ROUGE1: 38.8131  ROUGE2: 16.0827  ROUGEL: 24.6388  ROUGELSUM: 35.9153  GEN_LEN: 8126165","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Intel","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","whisper","whisper","Offline","5539.77","ACCURACY: 97.45763518206644","1","Intel(R) Xeon(R) 698X","1","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Intel/results/1-node-4x-BMG-B70/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Inventec","available","closed","datacenter","Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","P5800G7-B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","70269.6","exact_match: 81.22151321786691  TOKENS_PER_SAMPLE: 3748.471741112124","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P5800G7-B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","P5800G7-B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","57432.0","exact_match: 81.10756608933455  TOKENS_PER_SAMPLE: 3769.7967183226983","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P5800G7-B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","P5800G7-B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","106280.0","exact_match: 83.856","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P5800G7-B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","P5800G7-B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","100299.0","exact_match: 83.423","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P5800G7-B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","P5800G7-B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","109933.0","ROUGE1: 44.7871  ROUGE2: 22.3677  ROUGEL: 29.1429  TOKENS_PER_SAMPLE: 272.9","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P5800G7-B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","P5800G7-B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106617.0","ROUGE1: 44.8264  ROUGE2: 22.3815  ROUGEL: 29.1549  TOKENS_PER_SAMPLE: 272.1","1","Intel(R) Xeon(R) 6787P","2","86","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P5800G7-B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","P9000G7-B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","69062.2","exact_match: 81.13035551504102  TOKENS_PER_SAMPLE: 3743.316089334549","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P9000G7-B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","P9000G7-B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","57517.4","exact_match: 80.62898814949864  TOKENS_PER_SAMPLE: 3774.6118960802187","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P9000G7-B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","P9000G7-B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","105617.0","exact_match: 83.550","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P9000G7-B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","P9000G7-B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","100300.0","exact_match: 83.799","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P9000G7-B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","P9000G7-B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","109755.0","ROUGE1: 44.7774  ROUGE2: 22.365  ROUGEL: 29.1457  TOKENS_PER_SAMPLE: 272.9","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P9000G7-B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Inventec","available","closed","datacenter","Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","P9000G7-B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106645.0","ROUGE1: 44.8297  ROUGE2: 22.3857  ROUGEL: 29.1786  TOKENS_PER_SAMPLE: 272.2","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Inventec/results/P9000G7-B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Lambda","available","closed","datacenter","Lambda Cloud 8x B200-SXM-180GB","B200-SXM-180GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","101.56072395081954","0.7879966032452779","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B200-SXM-180GB","8","8","./closed/Lambda/results/B200-SXM-180GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT-LLM (GPT-OSS-120B); vLLM + NVIDIA Dynamo (Qwen3-VL-235B); CUDA 13.0","Ubuntu 24.04.4 LTS","8x NVIDIA B200-SXM-180GB (TGP 1000W).. GPT-OSS-120B via TensorRT-LLM LoadGen harness; Qwen3-VL-235B via NVIDIA inference-endpoint (vLLM+Dynamo), endpoint result format.","closed","0","v6.1","1","False","Samples/s","nvfp4","None",""
"Lambda","available","closed","datacenter","Lambda Cloud 8x B200-SXM-180GB","B200-SXM-180GBx8_TRT","qwen3-vl-235b-a22b","VLM","Server","69.41536196217709","0.7878950642363143","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B200-SXM-180GB","8","8","./closed/Lambda/results/B200-SXM-180GBx8_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT-LLM (GPT-OSS-120B); vLLM + NVIDIA Dynamo (Qwen3-VL-235B); CUDA 13.0","Ubuntu 24.04.4 LTS","8x NVIDIA B200-SXM-180GB (TGP 1000W).. GPT-OSS-120B via TensorRT-LLM LoadGen harness; Qwen3-VL-235B via NVIDIA inference-endpoint (vLLM+Dynamo), endpoint result format.","closed","0","v6.1","1","False","Queries/s","nvfp4","None",""
"Lambda","available","closed","datacenter","Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","GB300-NVL4_GB300-288GB_aarch64x4","gpt-oss-120b","gpt-oss-120b","Offline","65511.9","exact_match: 82.738","1","NVIDIA Grace CPU (Arm Neoverse-V2)","2","72","NVIDIA GB300","4","4","./closed/Lambda/results/GB300-NVL4_GB300-288GB_aarch64x4/gpt-oss-120b/Offline/performance/run_1","TensorRT-LLM 1.3.0rc14 (GPT-OSS-120B); vLLM (CentML) 0.20.2rc1.dev2091+g2e1040b7e + NVIDIA Dynamo (Qwen3-VL-235B); CUDA 13.0","Ubuntu 24.04.4 LTS (aarch64)","Single GB300 NVL4 node (Supermicro PIO-FRU-121GL-NB3-LCC-R1): 2x NVIDIA Grace CPU (72 cores each) + 4x NVIDIA GB300 288GB HBM3e, 1400W TGP per GPU, direct all-to-all 5th-gen NVLink (no NVLink Switch). GPU HBM is exposed as coherent NUMA memory alongside the 960GB of Grace LPDDR5X.. GPT-OSS-120B served with TensorRT-LLM (trtllm-serve, in-flight batching) as 4 data-parallel replicas (one GPU per replica) behind the MLPerf LoadGen harness. Qwen3-VL-235B-A22B served with vLLM + NVIDIA Dynamo as 4 data-parallel replicas (TP=1, one GPU per replica) and benchmarked with the NVIDIA/MLCommons inference-endpoints client; endpoint result format. Both workflows orchestrated by nv-sflow on Slurm with Pyxis/Enroot containers.","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"Lambda","available","closed","datacenter","Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","GB300-NVL4_GB300-288GB_aarch64x4","gpt-oss-120b","gpt-oss-120b","Server","58195.8","exact_match: 83.079","1","NVIDIA Grace CPU (Arm Neoverse-V2)","2","72","NVIDIA GB300","4","4","./closed/Lambda/results/GB300-NVL4_GB300-288GB_aarch64x4/gpt-oss-120b/Server/performance/run_1","TensorRT-LLM 1.3.0rc14 (GPT-OSS-120B); vLLM (CentML) 0.20.2rc1.dev2091+g2e1040b7e + NVIDIA Dynamo (Qwen3-VL-235B); CUDA 13.0","Ubuntu 24.04.4 LTS (aarch64)","Single GB300 NVL4 node (Supermicro PIO-FRU-121GL-NB3-LCC-R1): 2x NVIDIA Grace CPU (72 cores each) + 4x NVIDIA GB300 288GB HBM3e, 1400W TGP per GPU, direct all-to-all 5th-gen NVLink (no NVLink Switch). GPU HBM is exposed as coherent NUMA memory alongside the 960GB of Grace LPDDR5X.. GPT-OSS-120B served with TensorRT-LLM (trtllm-serve, in-flight batching) as 4 data-parallel replicas (one GPU per replica) behind the MLPerf LoadGen harness. Qwen3-VL-235B-A22B served with vLLM + NVIDIA Dynamo as 4 data-parallel replicas (TP=1, one GPU per replica) and benchmarked with the NVIDIA/MLCommons inference-endpoints client; endpoint result format. Both workflows orchestrated by nv-sflow on Slurm with Pyxis/Enroot containers.","closed","0","v6.1","0","False","Tokens/s","mxfp4","None",""
"Lambda","available","closed","datacenter","Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","GB300-NVL4_GB300-288GB_aarch64x4","qwen3-vl-235b-a22b","VLM","Offline","71.1217334486934","0.7874268401845349","1","NVIDIA Grace CPU (Arm Neoverse-V2)","2","72","NVIDIA GB300","4","4","./closed/Lambda/results/GB300-NVL4_GB300-288GB_aarch64x4/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT-LLM 1.3.0rc14 (GPT-OSS-120B); vLLM (CentML) 0.20.2rc1.dev2091+g2e1040b7e + NVIDIA Dynamo (Qwen3-VL-235B); CUDA 13.0","Ubuntu 24.04.4 LTS (aarch64)","Single GB300 NVL4 node (Supermicro PIO-FRU-121GL-NB3-LCC-R1): 2x NVIDIA Grace CPU (72 cores each) + 4x NVIDIA GB300 288GB HBM3e, 1400W TGP per GPU, direct all-to-all 5th-gen NVLink (no NVLink Switch). GPU HBM is exposed as coherent NUMA memory alongside the 960GB of Grace LPDDR5X.. GPT-OSS-120B served with TensorRT-LLM (trtllm-serve, in-flight batching) as 4 data-parallel replicas (one GPU per replica) behind the MLPerf LoadGen harness. Qwen3-VL-235B-A22B served with vLLM + NVIDIA Dynamo as 4 data-parallel replicas (TP=1, one GPU per replica) and benchmarked with the NVIDIA/MLCommons inference-endpoints client; endpoint result format. Both workflows orchestrated by nv-sflow on Slurm with Pyxis/Enroot containers.","closed","0","v6.1","1","False","Samples/s","nvfp4","None",""
"Lambda","available","closed","datacenter","Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","GB300-NVL4_GB300-288GB_aarch64x4","qwen3-vl-235b-a22b","VLM","Server","64.62448277938189","0.787136982780776","1","NVIDIA Grace CPU (Arm Neoverse-V2)","2","72","NVIDIA GB300","4","4","./closed/Lambda/results/GB300-NVL4_GB300-288GB_aarch64x4/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT-LLM 1.3.0rc14 (GPT-OSS-120B); vLLM (CentML) 0.20.2rc1.dev2091+g2e1040b7e + NVIDIA Dynamo (Qwen3-VL-235B); CUDA 13.0","Ubuntu 24.04.4 LTS (aarch64)","Single GB300 NVL4 node (Supermicro PIO-FRU-121GL-NB3-LCC-R1): 2x NVIDIA Grace CPU (72 cores each) + 4x NVIDIA GB300 288GB HBM3e, 1400W TGP per GPU, direct all-to-all 5th-gen NVLink (no NVLink Switch). GPU HBM is exposed as coherent NUMA memory alongside the 960GB of Grace LPDDR5X.. GPT-OSS-120B served with TensorRT-LLM (trtllm-serve, in-flight batching) as 4 data-parallel replicas (one GPU per replica) behind the MLPerf LoadGen harness. Qwen3-VL-235B-A22B served with vLLM + NVIDIA Dynamo as 4 data-parallel replicas (TP=1, one GPU per replica) and benchmarked with the NVIDIA/MLCommons inference-endpoints client; endpoint result format. Both workflows orchestrated by nv-sflow on Slurm with Pyxis/Enroot containers.","closed","0","v6.1","1","False","Queries/s","nvfp4","None",""
"MangoBoost","available","closed","datacenter","MangoBoost 2x8 MI355X (16x MI355X, LLMBoost P/D Disaggregated Interactive Engine)","16xMI355X_2xEPYC_9575F","gpt-oss-120b","gpt-oss-120b","Interactive","59297.4","exact_match: 83.867","2","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","16","./closed/MangoBoost/results/16xMI355X_2xEPYC_9575F/gpt-oss-120b/Interactive/performance/run_1","ROCm 7.2.0","Ubuntu 22.04.5 LTS (jammy), kernel 6.8.0-84-generic","2-node MI355X connect with 8x 40GB/s RDMA NICs. P/D Disaggregation for Interavtive Scenario, and vLLM 0.22.1","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"MangoBoost","available","closed","datacenter","MangoBoost 2x8 MI355X (16x MI355X, LLMBoost P/D Disaggregated Interactive Engine)","16xMI355X_2xEPYC_9575F","gpt-oss-120b","gpt-oss-120b","Offline","228488.0","exact_match: 83.502","2","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","16","./closed/MangoBoost/results/16xMI355X_2xEPYC_9575F/gpt-oss-120b/Offline/performance/run_1","ROCm 7.2.0","Ubuntu 22.04.5 LTS (jammy), kernel 6.8.0-84-generic","2-node MI355X connect with 8x 40GB/s RDMA NICs. P/D Disaggregation for Interavtive Scenario, and vLLM 0.22.1","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MangoBoost","available","closed","datacenter","8xMI300X_2xEPYC_9534","8xMI300X_2xEPYC_9534","gpt-oss-120b","gpt-oss-120b","Offline","30198.4","exact_match: 83.713","1","AMD EPYC 9534","2","64","AMD Instinct MI300X 192GB HBM3","8","8","./closed/MangoBoost/results/8xMI300X_2xEPYC_9534/gpt-oss-120b/Offline/performance/run_1","vLLM 0.24.0 (stock vllm/vllm-openai-rocm:v0.24.0), ROCm 7.2.1","Ubuntu 22.04.5 LTS","8xMI300X_2xEPYC_9534. Self-quantized w-fp8-a-fp8 gpt-oss-120b (AMD Quark 0.12) on stock vLLM 0.24.0 (vllm/vllm-openai-rocm:v0.24.0).","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MangoBoost","available","closed","datacenter","8xMI300X_2xEPYC_9534","8xMI300X_2xEPYC_9534","gpt-oss-120b","gpt-oss-120b","Server","24863.0","exact_match: 82.842","1","AMD EPYC 9534","2","64","AMD Instinct MI300X 192GB HBM3","8","8","./closed/MangoBoost/results/8xMI300X_2xEPYC_9534/gpt-oss-120b/Server/performance/run_1","vLLM 0.24.0 (stock vllm/vllm-openai-rocm:v0.24.0), ROCm 7.2.1","Ubuntu 22.04.5 LTS","8xMI300X_2xEPYC_9534. Self-quantized w-fp8-a-fp8 gpt-oss-120b (AMD Quark 0.12) on stock vLLM 0.24.0 (vllm/vllm-openai-rocm:v0.24.0).","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","gpt-oss-120b","gpt-oss-120b","Offline","97019.7","exact_match: 83.639","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/gpt-oss-120b/Offline/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","gpt-oss-120b","gpt-oss-120b","Server","87549.5","exact_match: 83.316","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/gpt-oss-120b/Server/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","llama2-70b-99.9","llama2-70b-99.9","Interactive","56673.0","ROUGE1: 44.5133  ROUGE2: 22.148  ROUGEL: 28.9008  TOKENS_PER_SAMPLE: 284.9","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/llama2-70b-99.9/Interactive/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","llama2-70b-99.9","llama2-70b-99.9","Offline","84016.4","ROUGE1: 44.4551  ROUGE2: 22.0787  ROUGEL: 28.8207  TOKENS_PER_SAMPLE: 283.0","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/llama2-70b-99.9/Offline/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","llama2-70b-99.9","llama2-70b-99.9","Server","85004.8","ROUGE1: 44.5423  ROUGE2: 22.1706  ROUGEL: 28.9236  TOKENS_PER_SAMPLE: 284.5","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/llama2-70b-99.9/Server/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","llama3.1-8b","llama3.1-8b","Interactive","104368.0","ROUGE1: 38.5707  ROUGE2: 16.0354  ROUGEL: 24.4433  ROUGELSUM: 35.708  GEN_LEN: 8150585","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/llama3.1-8b/Interactive/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","llama3.1-8b","llama3.1-8b","Offline","129943.0","ROUGE1: 38.5694  ROUGE2: 16.0367  ROUGEL: 24.4456  ROUGELSUM: 35.7078  GEN_LEN: 8150329","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/llama3.1-8b/Offline/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","G8825Z5","8xMI350X_2xEPYC-9755","llama3.1-8b","llama3.1-8b","Server","121706.0","ROUGE1: 38.5208  ROUGE2: 16.0104  ROUGEL: 24.424  ROUGELSUM: 35.666  GEN_LEN: 8168845","1","AMD EPYC 9755","2","128","AMD Instinct MI350X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI350X_2xEPYC-9755/llama3.1-8b/Server/performance/run_1","Pytorch 2.10, ROCm 7.2.2","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","gpt-oss-120b","gpt-oss-120b","Offline","118042.0","exact_match: 83.861","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","gpt-oss-120b","gpt-oss-120b","Server","108861.0","exact_match: 84.039","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","llama2-70b-99.9","llama2-70b-99.9","Interactive","74385.6","ROUGE1: 44.5012  ROUGE2: 22.1509  ROUGEL: 28.9111  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/llama2-70b-99.9/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","llama2-70b-99.9","llama2-70b-99.9","Offline","104904.0","ROUGE1: 44.5341  ROUGE2: 22.1837  ROUGEL: 28.9253  TOKENS_PER_SAMPLE: 284.3","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/llama2-70b-99.9/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","llama2-70b-99.9","llama2-70b-99.9","Server","104541.0","ROUGE1: 44.5342  ROUGE2: 22.1734  ROUGEL: 28.9337  TOKENS_PER_SAMPLE: 284.0","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/llama2-70b-99.9/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","llama3.1-8b","llama3.1-8b","Interactive","136999.0","ROUGE1: 38.5359  ROUGE2: 16.0141  ROUGEL: 24.4288  ROUGELSUM: 35.6751  GEN_LEN: 8162900","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/llama3.1-8b/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","llama3.1-8b","llama3.1-8b","Offline","159278.0","ROUGE1: 38.5554  ROUGE2: 16.0227  ROUGEL: 24.4331  ROUGELSUM: 35.6963  GEN_LEN: 8150276","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/llama3.1-8b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","G4826","8xMI355X_2xEPYC-9755","llama3.1-8b","llama3.1-8b","Server","151116.0","ROUGE1: 38.5121  ROUGE2: 15.9993  ROUGEL: 24.4128  ROUGELSUM: 35.6518  GEN_LEN: 8170944","1","AMD EPYC 9755","2","128","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/MiTAC/results/8xMI355X_2xEPYC-9755/llama3.1-8b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211","Ubuntu 24.04.4 LTS","","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"MiTAC","available","closed","datacenter","MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8","MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","30034.2","ROUGE1: 44.7186  ROUGE2: 22.3366  ROUGEL: 29.1057  TOKENS_PER_SAMPLE: 275.5","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/MiTAC/results/MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.2","Ubuntu 24.04.4","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"MiTAC","available","closed","datacenter","MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8","MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","29203.3","ROUGE1: 44.8047  ROUGE2: 22.3347  ROUGEL: 29.1315  TOKENS_PER_SAMPLE: 267.8","1","Intel(R) Xeon(R) 6787P","2","172","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/MiTAC/results/MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.2","Ubuntu 24.04.4","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","70132.3","exact_match: 80.78851412944394  TOKENS_PER_SAMPLE: 3793.9359617137648","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","59867.8","exact_match: 81.17593436645396  TOKENS_PER_SAMPLE: 3818.7203737465816","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","106637.0","exact_match: 83.430","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","100630.0","exact_match: 83.702","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","109277.0","ROUGE1: 44.7951  ROUGE2: 22.3759  ROUGEL: 29.1534  TOKENS_PER_SAMPLE: 272.8","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106949.0","ROUGE1: 44.8452  ROUGE2: 22.4177  ROUGEL: 29.2149  TOKENS_PER_SAMPLE: 272.7","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Interactive","3.9944289632075423","0.7863529411764706","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","bf16","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","129.35839172137014","0.788067250804375","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","bf16","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Server","104.57819163892032","0.787266660240787","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","bf16","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.062188662827408765","0.7024464645142161","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"NVIDIA","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","17.97718330692361","0.7019984343473832","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/NVIDIA/results/B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Interactive","240274.0","exact_match: 81.22151321786691  TOKENS_PER_SAMPLE: 3834.577939835916","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","523790.0","exact_match: 81.0619872379216  TOKENS_PER_SAMPLE: 3795.436417502279","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","395940.0","exact_match: 80.78851412944394  TOKENS_PER_SAMPLE: 3823.4881494986325","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Interactive","625205.0","exact_match: 83.374","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/gpt-oss-120b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Offline","883761.0","exact_match: 84.038","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Server","899005.0","exact_match: 83.914","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.4960942571652712","0.7018512713495823","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","GB200-NVL72_GB200-186GB_aarch64x72_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","5.731549943319444","0.7027881259253183","18","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","72","./closed/NVIDIA/results/GB200-NVL72_GB200-186GB_aarch64x72_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB200 NVL72, 1200W TGP Per GPU","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x288_TRT","deepseek-r1","deepseek-r1","Offline","2705130.0","exact_match: 81.19872379216044  TOKENS_PER_SAMPLE: 3809.0271194165907","72","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","288","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x288_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x288_TRT","deepseek-r1","deepseek-r1","Server","2028030.0","exact_match: 81.13035551504102  TOKENS_PER_SAMPLE: 3861.4316317228804","72","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","288","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x288_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Interactive","253506.0","exact_match: 81.40382862351869  TOKENS_PER_SAMPLE: 3825.619416590702","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","679740.0","exact_match: 80.9936189608022  TOKENS_PER_SAMPLE: 3800.5963992707384","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","596944.0","exact_match: 81.42661804922516  TOKENS_PER_SAMPLE: 3855.550136736554","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Interactive","669305.0","exact_match: 83.495","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Offline","1124500.0","exact_match: 82.861","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Server","1120710.0","exact_match: 83.069","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Interactive","349.2921988473666","0.7853930563050908","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","bf16","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Offline","1304.9852625797478","0.787383322355058","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","bf16","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Server","1210.48823482974","0.7869294905778593","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","bf16","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.6546111599325777","0.6979912810173431","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","5.807651162902777","0.6998451039205666","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","deepseek-r1","deepseek-r1","Offline","75918.2","exact_match: 81.13035551504102  TOKENS_PER_SAMPLE: 3843.4847310847767","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","deepseek-r1","deepseek-r1","Server","67578.3","exact_match: 80.78851412944394  TOKENS_PER_SAMPLE: 3829.014357338195","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","131619.0","exact_match: 82.963","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","gpt-oss-120b","gpt-oss-120b","Server","128173.0","exact_match: 83.020","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","131464.0","ROUGE1: 44.8017  ROUGE2: 22.3794  ROUGEL: 29.1616  TOKENS_PER_SAMPLE: 272.6","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","127911.0","ROUGE1: 44.8324  ROUGE2: 22.3953  ROUGEL: 29.1865  TOKENS_PER_SAMPLE: 272.1","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/NVIDIA/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.1","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"NVIDIA","available","closed","edge","NVIDIA Jetson AGX Thor Developer Kit 128G (TensorRT)","Jetson-Thor","qwen3.6-27b","agentic edge","SingleStream","1466.007675716981","0.8793969849246231","1","14-core ARM Poseidon-AE CPU","1","14","NVIDIA Jetson AGX Thor 128G","1","1","./closed/NVIDIA/results/Jetson-Thor/qwen3.6-27b/SingleStream/performance/run_1","TensorRT 10.16, CUDA 13.2","Jetson L4T","","closed","0","v6.1","1","False","Latency (ms)","fp4","None",""
"NVIDIA","preview","closed","datacenter","NVIDIA VR200 NVL72 (72x VR200-288GB_aarch64)","VR200-NVL72_GR200-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Interactive","652750.1447708113","80.62898814949864","18","NVIDIA Vera CPU","2","144","NVIDIA VR200","4","72","./closed/NVIDIA/results/VR200-NVL72_GR200-288GB_aarch64x72_TRT/deepseek-r1/Interactive/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","preview","closed","datacenter","NVIDIA VR200 NVL72 (72x VR200-288GB_aarch64)","VR200-NVL72_GR200-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","1183326.8540322022","81.35824977210574","18","NVIDIA Vera CPU","2","144","NVIDIA VR200","4","72","./closed/NVIDIA/results/VR200-NVL72_GR200-288GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","preview","closed","datacenter","NVIDIA VR200 NVL72 (72x VR200-288GB_aarch64)","VR200-NVL72_GR200-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","1175890.2208168241","81.42661804922516","18","NVIDIA Vera CPU","2","144","NVIDIA VR200","4","72","./closed/NVIDIA/results/VR200-NVL72_GR200-288GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"NVIDIA","preview","closed","datacenter","NVIDIA VR200 NVL72 (72x VR200-288GB_aarch64)","VR200-NVL72_GR200-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Interactive","1306.6078446834506","0.7845128091628396","18","NVIDIA Vera CPU","2","144","NVIDIA VR200","4","72","./closed/NVIDIA/results/VR200-NVL72_GR200-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","bf16","None",""
"NVIDIA","preview","closed","datacenter","NVIDIA VR200 NVL72 (72x VR200-288GB_aarch64)","VR200-NVL72_GR200-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Offline","2392.6979194587702","0.7871815525981819","18","NVIDIA Vera CPU","2","144","NVIDIA VR200","4","72","./closed/NVIDIA/results/VR200-NVL72_GR200-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","bf16","None",""
"NVIDIA","preview","closed","datacenter","NVIDIA VR200 NVL72 (72x VR200-288GB_aarch64)","VR200-NVL72_GR200-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Server","2323.2698417825063","0.7874153796727632","18","NVIDIA Vera CPU","2","144","NVIDIA VR200","4","72","./closed/NVIDIA/results/VR200-NVL72_GR200-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","bf16","None",""
"Naeem Khoshnevis","rdi","closed","datacenter","H200-1xGPU","H200","llama3.1-8b","llama3.1-8b","Offline","7796.585132594509","rouge1: 38.6405  rouge2: 15.9282  rougeL: 24.5018  rougeLsum: 35.7399  gen_len: 8222392  gen_num: 13368","1","AMD EPYC 9374F 32-Core Processor","2","32","NVIDIA H200-SXM-141GB","1","1","./closed/Naeem Khoshnevis/results/H200/llama3.1-8b/Offline/performance/run_1","TensorRT-LLM 1.0.0 (FP8 W8A8 engine built with trtllm-build, served by trtllm-serve); MLCommons endpoints (LoadGen++) client; CUDA 12.9","Rocky Linux 8.10","H200 node; 4x NVIDIA H200 141GB per node, 1 GPU used for llama3.1-8b. Subset-of-node use is disclosed: the job was cgroup-limited to 16 host CPU cores and 220-240 GB of host RAM depending on scenario (Offline 240 GB, Server 220 GB) out of the node's 2x32-core CPUs and 1.5 TB. GPU ECC verified enabled via nvidia-smi on the GPU used.. llama3.1-8b served by TensorRT-LLM 1.0.0 as an FP8 (W8A8) engine with FP8 KV cache. Weights quantized by NVIDIA ModelOpt static per-tensor PTQ calibrated on the MLPerf-official CNN/DailyMail calibration id list (calibration/CNNDailyMail/calibration-list.txt, 998 unique ids; ModelOpt consumed a 512-row prefix spanning 510 unique ids). Calibration inputs were reconstructed from the ids as raw article text, not the reference instruction-templated calibration file; see documentation/calibration.md. TP1. KV-cache block reuse explicitly DISABLED via --extra_llm_api_options (enable_block_reuse, enable_partial_reuse and copy_on_partial_reuse all false in both scenarios).","closed","0","v6.1","1","False","Tokens/s","fp8 (static per-tensor PTQ, NVIDIA ModelOpt)","None",""
"Naeem Khoshnevis","rdi","closed","datacenter","H200-1xGPU","H200","llama3.1-8b","llama3.1-8b","Server","7376.597666232419","rouge1: 38.6354  rouge2: 15.9256  rougeL: 24.4991  rougeLsum: 35.7316  gen_len: 8222691  gen_num: 13368","1","AMD EPYC 9374F 32-Core Processor","2","32","NVIDIA H200-SXM-141GB","1","1","./closed/Naeem Khoshnevis/results/H200/llama3.1-8b/Server/performance/run_1","TensorRT-LLM 1.0.0 (FP8 W8A8 engine built with trtllm-build, served by trtllm-serve); MLCommons endpoints (LoadGen++) client; CUDA 12.9","Rocky Linux 8.10","H200 node; 4x NVIDIA H200 141GB per node, 1 GPU used for llama3.1-8b. Subset-of-node use is disclosed: the job was cgroup-limited to 16 host CPU cores and 220-240 GB of host RAM depending on scenario (Offline 240 GB, Server 220 GB) out of the node's 2x32-core CPUs and 1.5 TB. GPU ECC verified enabled via nvidia-smi on the GPU used.. llama3.1-8b served by TensorRT-LLM 1.0.0 as an FP8 (W8A8) engine with FP8 KV cache. Weights quantized by NVIDIA ModelOpt static per-tensor PTQ calibrated on the MLPerf-official CNN/DailyMail calibration id list (calibration/CNNDailyMail/calibration-list.txt, 998 unique ids; ModelOpt consumed a 512-row prefix spanning 510 unique ids). Calibration inputs were reconstructed from the ids as raw article text, not the reference instruction-templated calibration file; see documentation/calibration.md. TP1. KV-cache block reuse explicitly DISABLED via --extra_llm_api_options (enable_block_reuse, enable_partial_reuse and copy_on_partial_reuse all false in both scenarios).","closed","0","v6.1","1","False","Tokens/s","fp8 (static per-tensor PTQ, NVIDIA ModelOpt)","None",""
"Nebius","available","closed","datacenter","Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","B200-SXM-180GBx8_TRT","deepseek-r1","deepseek-r1","Offline","56929.9","exact_match: 81.08477666362808  TOKENS_PER_SAMPLE: 3808.5396536007293","1","Intel(R) Xeon(R) Platinum 8580","2","60","NVIDIA B200-SXM-180GB","8","8","./closed/Nebius/results/B200-SXM-180GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04.3 LTS (Linux kernel: 6.11.0-1016-nvidia #16-Ubuntu SMP PREEMPT_DYNAMIC Sun Sep 21 17:06:33 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux)","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","B200-SXM-180GBx8_TRT","deepseek-r1","deepseek-r1","Server","55869.0","exact_match: 80.62898814949864  TOKENS_PER_SAMPLE: 3840.4710574293526","1","Intel(R) Xeon(R) Platinum 8580","2","60","NVIDIA B200-SXM-180GB","8","8","./closed/Nebius/results/B200-SXM-180GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04.3 LTS (Linux kernel: 6.11.0-1016-nvidia #16-Ubuntu SMP PREEMPT_DYNAMIC Sun Sep 21 17:06:33 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux)","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","B200-SXM-180GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","88491.0","exact_match: 83.040","1","Intel(R) Xeon(R) Platinum 8580","2","60","NVIDIA B200-SXM-180GB","8","8","./closed/Nebius/results/B200-SXM-180GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04.3 LTS (Linux kernel: 6.11.0-1016-nvidia #16-Ubuntu SMP PREEMPT_DYNAMIC Sun Sep 21 17:06:33 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux)","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","B200-SXM-180GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","89856.3","exact_match: 82.998","1","Intel(R) Xeon(R) Platinum 8580","2","60","NVIDIA B200-SXM-180GB","8","8","./closed/Nebius/results/B200-SXM-180GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04.3 LTS (Linux kernel: 6.11.0-1016-nvidia #16-Ubuntu SMP PREEMPT_DYNAMIC Sun Sep 21 17:06:33 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux)","B200 TGP 1000W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","69655.1","exact_match: 80.81130355515042  TOKENS_PER_SAMPLE: 3831.6855059252507","1","Intel(R) Xeon(R) Platinum 8570","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/Nebius/results/B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","63909.3","exact_match: 81.67730173199635  TOKENS_PER_SAMPLE: 3826.9986326344574","1","Intel(R) Xeon(R) Platinum 8570","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/Nebius/results/B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","108529.0","exact_match: 83.782","1","Intel(R) Xeon(R) Platinum 8570","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/Nebius/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","108124.0","exact_match: 83.224","1","Intel(R) Xeon(R) Platinum 8570","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/Nebius/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","131.47716300755746","0.7877591584823119","1","Intel(R) Xeon(R) Platinum 8570","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/Nebius/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","BF16/F32/F8_E4M3/U8","None",""
"Nebius","available","closed","datacenter","Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Server","112.28582371293707","0.7876168551420853","1","Intel(R) Xeon(R) Platinum 8570","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/Nebius/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","BF16/F32/F8_E4M3/U8","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Offline","73.18605442300793","0.7874581413977274","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","nvfp4, fp8 KV cache","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Server","64.61995222586359","0.7874065982485451","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","nvfp4, fp8 KV cache","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","689961.0","exact_match: 80.94804010938924  TOKENS_PER_SAMPLE: 3831.7297174111213","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","603023.0","exact_match: 80.67456700091158  TOKENS_PER_SAMPLE: 3850.7272105742936","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Offline","1186750.0","exact_match: 82.905","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Server","1122490.0","exact_match: 83.174","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","132236.0","exact_match: 82.975","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","gpt-oss-120b","gpt-oss-120b","Server","127921.0","exact_match: 83.209","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/Nebius/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius RTX PRO 6000 n1 (8x RTX-PRO-6000-PCIE-96GB, TensorRT)","RTX_PRO_6000_PCIE_96GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","15738.5","exact_match: 83.832","1","Intel(R) Xeon(R) 6776P","2","96","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Nebius/results/RTX_PRO_6000_PCIE_96GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04.4 LTS, Linux 6.11.0-1016-nvidia","KVM virtual machine with 192 vCPUs presented as 2 sockets x 48 cores x 2 threads; 2 virtual NUMA nodes; 8 attached GPUs with 600 W power limit each. nvidia-smi reports CUDA 13.0 driver compatibility; CUDA toolkit and runtime versions are container-specific","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","available","closed","datacenter","Nebius RTX PRO 6000 n1 (8x RTX-PRO-6000-PCIE-96GB, TensorRT)","RTX_PRO_6000_PCIE_96GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","14784.3","exact_match: 83.717","1","Intel(R) Xeon(R) 6776P","2","96","NVIDIA RTX PRO 6000 Blackwell Server Edition","8","8","./closed/Nebius/results/RTX_PRO_6000_PCIE_96GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04.4 LTS, Linux 6.11.0-1016-nvidia","KVM virtual machine with 192 vCPUs presented as 2 sockets x 48 cores x 2 threads; 2 virtual NUMA nodes; 8 attached GPUs with 600 W power limit each. nvidia-smi reports CUDA 13.0 driver compatibility; CUDA toolkit and runtime versions are container-specific","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Nebius","preview","closed","datacenter","Nebius VR200 NVL72 (36x VR200-288GB_aarch64, TensorRT)","VR200-NVL72_aarch64x36","deepseek-r1","deepseek-r1","Offline","558802.8293934778","81.56335460346399","9","NVIDIA Vera CPU","2","352","NVIDIA VR200","4","36","./closed/Nebius/results/VR200-NVL72_aarch64x36/deepseek-r1/Offline/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Nebius","preview","closed","datacenter","Nebius VR200 NVL72 (36x VR200-288GB_aarch64, TensorRT)","VR200-NVL72_aarch64x36","deepseek-r1","deepseek-r1","Server","591368.312764167","81.03919781221514","9","NVIDIA Vera CPU","2","352","NVIDIA VR200","4","36","./closed/Nebius/results/VR200-NVL72_aarch64x36/deepseek-r1/Server/performance/run_1","TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo","Ubuntu 24.04","NVIDIA VR200 NVL72, 2300W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","BM.Standard4.Ax.120","1-node-1S-GNR_120C","llama3_1-8b","llama3.1-8b","Offline","853.14","ROUGE1: 38.7451  ROUGE2: 16.0022  ROUGEL: 24.5948  ROUGELSUM: 35.807  GEN_LEN: 8190646","1","Intel(R) Xeon(R) 6987P-C","1","120","N/A","0","0","./closed/Oracle/results/1-node-1S-GNR_120C/llama3_1-8b/Offline/performance/run_1","PyTorch","Oracle-Linux-9.7-2026.06.15-1","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Oracle","available","closed","datacenter","BM.Standard4.Ax.120","1-node-1S-GNR_120C","llama3_1-8b","llama3.1-8b","Server","324.995","ROUGE1: 38.7167  ROUGE2: 16.0067  ROUGEL: 24.5623  ROUGELSUM: 35.7763  GEN_LEN: 8199230","1","Intel(R) Xeon(R) 6987P-C","1","120","N/A","0","0","./closed/Oracle/results/1-node-1S-GNR_120C/llama3_1-8b/Server/performance/run_1","PyTorch","Oracle-Linux-9.7-2026.06.15-1","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Oracle","available","closed","datacenter","BM.Standard4.Ax.120","1-node-1S-GNR_120C","whisper","whisper","Offline","1099.64","ACCURACY: 97.86360951548667","1","Intel(R) Xeon(R) 6987P-C","1","120","N/A","0","0","./closed/Oracle/results/1-node-1S-GNR_120C/whisper/Offline/performance/run_1","PyTorch","Oracle-Linux-9.7-2026.06.15-1","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","gpt-oss-120b","gpt-oss-120b","Offline","115567.0","exact_match: 83.895","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","gpt-oss-120b","gpt-oss-120b","Server","109083.0","exact_match: 83.679","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","llama2-70b-99.9","llama2-70b-99.9","Interactive","70421.7","ROUGE1: 44.5013  ROUGE2: 22.1522  ROUGEL: 28.9162  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/llama2-70b-99.9/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","llama2-70b-99.9","llama2-70b-99.9","Offline","101553.0","ROUGE1: 44.5168  ROUGE2: 22.1764  ROUGEL: 28.9174  TOKENS_PER_SAMPLE: 284.5","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/llama2-70b-99.9/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","llama2-70b-99.9","llama2-70b-99.9","Server","100392.0","ROUGE1: 44.5335  ROUGE2: 22.1719  ROUGEL: 28.9333  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/llama2-70b-99.9/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","llama3.1-8b","llama3.1-8b","Interactive","130625.0","ROUGE1: 38.5651  ROUGE2: 16.0381  ROUGEL: 24.4433  ROUGELSUM: 35.7047  GEN_LEN: 8150907","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/llama3.1-8b/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","1","False","Tokens/s","fp8","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","llama3.1-8b","llama3.1-8b","Offline","144321.0","ROUGE1: 38.5632  ROUGE2: 16.0256  ROUGEL: 24.4374  ROUGELSUM: 35.7022  GEN_LEN: 8150941","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/llama3.1-8b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","llama3.1-8b","llama3.1-8b","Server","148594.0","ROUGE1: 38.5373  ROUGE2: 16.018  ROUGEL: 24.4335  ROUGELSUM: 35.6767  GEN_LEN: 8166290","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/llama3.1-8b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","wan-2.2-t2v-a14b","text_to_video","Offline","0.0706215","vbench_score: 69.9515","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/wan-2.2-t2v-a14b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Samples/s","bf16, fp8, fp4","None",""
"Oracle","available","closed","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC-9575F","wan-2.2-t2v-a14b","text_to_video","SingleStream","16.334597792","vbench_score: 70.3415","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Oracle/results/8xMI355X_2xEPYC-9575F/wan-2.2-t2v-a14b/SingleStream/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2","Ubuntu 22.04.2 LTS (kernel 5.15.0-70-generic)","8x AMD Instinct MI355X (gfx950), 288GB HBM3E each. MXFP4 (W4A4) weights with FP8 KV cache, Quark quantization","closed","0","v6.1","0","False","Latency (s)","bf16, fp8, fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","70354.0","exact_match: 80.9936189608022  TOKENS_PER_SAMPLE: 3810.1563354603463","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","60745.5","exact_match: 81.35824977210574  TOKENS_PER_SAMPLE: 3858.553099361896","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","109931.0","exact_match: 83.311","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","100608.0","exact_match: 83.635","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","108026.0","ROUGE1: 44.7932  ROUGE2: 22.3753  ROUGEL: 29.152  TOKENS_PER_SAMPLE: 272.9","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106696.0","ROUGE1: 44.8216  ROUGE2: 22.385  ROUGEL: 29.1799  TOKENS_PER_SAMPLE: 272.5","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Interactive","3.9944621629556996","0.7876188235945263","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","124.38597378730044","0.7878245382101281","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","8","8","./closed/Oracle/results/B300-SXM-270GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","gpt-oss-120b","gpt-oss-120b","Offline","60745.8","exact_match: 83.395","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","gpt-oss-120b","gpt-oss-120b","Server","59996.0","exact_match: 83.331","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","66001.9","ROUGE1: 44.7161  ROUGE2: 22.308  ROUGEL: 29.1704  TOKENS_PER_SAMPLE: 267.0","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","58991.9","ROUGE1: 44.7161  ROUGE2: 22.308  ROUGEL: 29.1704  TOKENS_PER_SAMPLE: 267.0","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Interactive","4.992209807647115","0.7862706307153826","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Offline","72.78618281809769","0.7881223292682367","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","qwen3-vl-235b-a22b","VLM","Server","64.60782589940354","0.7881283191730583","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.013837221211163476","0.7034961544029444","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","30.653359842875002","0.699114664501586","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Interactive","235807.0","exact_match: 81.28988149498633  TOKENS_PER_SAMPLE: 3770.5820419325432","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Interactive/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Offline","666613.0","exact_match: 81.42661804922516  TOKENS_PER_SAMPLE: 3825.77484047402","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","deepseek-r1","deepseek-r1","Server","586852.0","exact_match: 81.38103919781221  TOKENS_PER_SAMPLE: 3780.2771194165907","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Offline","1097620.0","exact_match: 83.395","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","gpt-oss-120b","gpt-oss-120b","Server","1092610.0","exact_match: 83.331","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Offline","1297.4909412305838","0.7868281285145724","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","qwen3-vl-235b-a22b","VLM","Server","1146.388942892591","0.7878666850296102","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Queries/s","NVFP4 weights and activations; FP8 KV cache","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.3997628673298144","0.6948773298873472","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Samples/s","fp8","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x72_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","5.895425553486111","0.7028506315560222","18","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","72","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x72_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","1","False","Latency (s)","fp8","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","deepseek-r1","deepseek-r1","Offline","77031.7","exact_match: 81.42661804922516  TOKENS_PER_SAMPLE: 3803.590474020055","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","deepseek-r1","deepseek-r1","Server","68138.0","exact_match: 80.9024612579763  TOKENS_PER_SAMPLE: 3861.967183226983","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","132253.0","ROUGE1: 44.8005  ROUGE2: 22.3791  ROUGEL: 29.1527  TOKENS_PER_SAMPLE: 272.6","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Oracle","available","closed","datacenter","NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","127984.0","ROUGE1: 44.8018  ROUGE2: 22.3668  ROUGEL: 29.141  TOKENS_PER_SAMPLE: 272.9","2","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","8","./closed/Oracle/results/GB300-NVL72_GB300-288GB_aarch64x8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","llama3_1-8b","llama3.1-8b","Offline","1170.51","ROUGE1: 38.703  ROUGE2: 15.9823  ROUGEL: 24.5421  ROUGELSUM: 35.7696  GEN_LEN: 8188918","1","Intel(R) Xeon(R) 6788P","2","86","N/A","0","0","./closed/Quanta_Cloud_Technology/results/1-node-2S-GNR_86C/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","llama3_1-8b","llama3.1-8b","Server","516.012","ROUGE1: 38.6189  ROUGE2: 15.8198  ROUGEL: 24.4192  ROUGELSUM: 35.6324  GEN_LEN: 8192507","1","Intel(R) Xeon(R) 6788P","2","86","N/A","0","0","./closed/Quanta_Cloud_Technology/results/1-node-2S-GNR_86C/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","rgat","rgat","Offline","15633.3","acc: 71.489","1","Intel(R) Xeon(R) 6788P","2","86","N/A","0","0","./closed/Quanta_Cloud_Technology/results/1-node-2S-GNR_86C/rgat/Offline/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","whisper","whisper","Offline","1548.1","ACCURACY: 97.89045996917045","1","Intel(R) Xeon(R) 6788P","2","86","N/A","0","0","./closed/Quanta_Cloud_Technology/results/1-node-2S-GNR_86C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama3_1-8b","llama3.1-8b","Offline","4977.67","ROUGE1: 38.8427  ROUGE2: 16.0895  ROUGEL: 24.6432  ROUGELSUM: 35.9449  GEN_LEN: 8127665","1","Intel(R) Xeon(R) 6788P","2","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Quanta_Cloud_Technology/results/1-node-4x-BMG-B70/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama3_1-8b","llama3.1-8b","Server","4122.73","ROUGE1: 38.8436  ROUGE2: 16.0905  ROUGEL: 24.6447  ROUGELSUM: 35.9462  GEN_LEN: 8127418","1","Intel(R) Xeon(R) 6788P","2","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Quanta_Cloud_Technology/results/1-node-4x-BMG-B70/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","whisper","whisper","Offline","5401.23","ACCURACY: 97.4509990485252","1","Intel(R) Xeon(R) 6788P","2","86","Intel(R) Arc(TM) Pro B70","4","4","./closed/Quanta_Cloud_Technology/results/1-node-4x-BMG-B70/whisper/Offline/performance/run_1","PyTorch","Ubuntu 25.10","QuantaGrid D75E-4U. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","69841.6","exact_match: 81.58614402917047  TOKENS_PER_SAMPLE: 3773.390154968095","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","59622.2","exact_match: 81.17593436645396  TOKENS_PER_SAMPLE: 3821.9033728350046","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Offline","108293.0","exact_match: 83.431","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","gpt-oss-120b","gpt-oss-120b","Server","100183.0","exact_match: 83.451","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","113985.0","ROUGE1: 44.7969  ROUGE2: 22.3741  ROUGEL: 29.152  TOKENS_PER_SAMPLE: 272.6","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106285.0","ROUGE1: 44.8968  ROUGE2: 22.4583  ROUGEL: 29.2397  TOKENS_PER_SAMPLE: 272.0","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Offline","131.7949949301423","0.7880292841897505","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","1","False","Samples/s","fp32","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","qwen3-vl-235b-a22b","VLM","Server","104.81182396432094","0.7875545886602456","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","1","False","Queries/s","fp32","None",""
"Quanta_Cloud_Technology","available","closed","datacenter","QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","D75H-10U_B300-270GBx8_TRT","whisper","whisper","Offline","45408.2","ACCURACY: 97.82350086127316","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Quanta_Cloud_Technology/results/D75H-10U_B300-270GBx8_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Rocky Linux 10.1","QuantaGrid D75H-10U","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"RedHat_Intel","available","closed","datacenter","1-node-2S-GNR_96C","1-node-2S-GNR_96C","llama3_1-8b","llama3.1-8b","Offline","1507.21","ROUGE1: 38.7872  ROUGE2: 16.0024  ROUGEL: 24.5918  ROUGELSUM: 35.7668  GEN_LEN: 8190420","1","Intel(R) Xeon(R) 6972P","2","96","N/A","0","0","./closed/RedHat_Intel/results/1-node-2S-GNR_96C/llama3_1-8b/Offline/performance/run_1","PyTorch","RHEL 9.9","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"RedHat_Intel","available","closed","datacenter","1-node-2S-GNR_96C","1-node-2S-GNR_96C","llama3_1-8b","llama3.1-8b","Server","644.788","ROUGE1: 38.606  ROUGE2: 15.7593  ROUGEL: 24.3946  ROUGELSUM: 35.6342  GEN_LEN: 8185295","1","Intel(R) Xeon(R) 6972P","2","96","N/A","0","0","./closed/RedHat_Intel/results/1-node-2S-GNR_96C/llama3_1-8b/Server/performance/run_1","PyTorch","RHEL 9.9","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"RedHat_Intel","available","closed","datacenter","1-node-2S-GNR_96C","1-node-2S-GNR_96C","whisper","whisper","Offline","1966.11","ACCURACY: 97.86193444026833","1","Intel(R) Xeon(R) 6972P","2","96","N/A","0","0","./closed/RedHat_Intel/results/1-node-2S-GNR_96C/whisper/Offline/performance/run_1","PyTorch","RHEL 9.9","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"RedHat_Supermicro","available","closed","datacenter","NVIDIA GB200 NVL4 (4x GB200-192GB_aarch64, vLLM)","GB200-NVL4-192GB_aarch64x4_VLLM","qwen3-vl-235b-a22b","VLM","Offline","58.07798401246531","0.7874682472480947","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/RedHat_Supermicro/results/GB200-NVL4-192GB_aarch64x4_VLLM/qwen3-vl-235b-a22b/Offline/performance/run_1","vLLM v0.20.2rc1.dev2091, CUDA 13.0.1, NVIDIA Dynamo, PyTorch 2.11.0+cu130","Red Hat Enterprise Linux 9.8","NVIDIA GB200 NVL4, 1200W TGP Per GPU, Supermicro chassis","closed","0","v6.1","1","False","Samples/s","fp4","None",""
"RedHat_Supermicro","available","closed","datacenter","NVIDIA GB200 NVL4 (4x GB200-192GB_aarch64, vLLM)","GB200-NVL4-192GB_aarch64x4_VLLM","qwen3-vl-235b-a22b","VLM","Server","50.683789589484135","0.7877985363200543","1","NVIDIA Grace CPU","2","144","NVIDIA GB200","4","4","./closed/RedHat_Supermicro/results/GB200-NVL4-192GB_aarch64x4_VLLM/qwen3-vl-235b-a22b/Server/performance/run_1","vLLM v0.20.2rc1.dev2091, CUDA 13.0.1, NVIDIA Dynamo, PyTorch 2.11.0+cu130","Red Hat Enterprise Linux 9.8","NVIDIA GB200 NVL4, 1200W TGP Per GPU, Supermicro chassis","closed","0","v6.1","1","False","Queries/s","fp4","None",""
"RedHat_Supermicro","available","closed","datacenter","SMC Lab OpenShift GB200-NVL4 4xGB200-186GB","GB200-NVL4_186GB_Openshift_VLLM","gpt-oss-120b","gpt-oss-120b","Offline","61205.6","exact_match: 83.695","1","NVIDIA Grace CPU","2","72","NVIDIA GB200","4","4","./closed/RedHat_Supermicro/results/GB200-NVL4_186GB_Openshift_VLLM/gpt-oss-120b/Offline/performance/run_1","vLLM 0.24.0, Red Hat OpenShift 4.21","Red Hat Enterprise Linux CoreOS 9.6","NVIDIA GB200 NVL4 with 2x Grace CPUs (72 ARM cores each), 4x GB200 GPUs (189GB HBM3e each). SMC Lab, Red Hat OpenShift 4.21, RHEL CoreOS 9.6, vLLM 0.24.0","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"RedHat_Supermicro","available","closed","datacenter","SMC Lab OpenShift GB200-NVL4 4xGB200-186GB","GB200-NVL4_186GB_Openshift_VLLM","gpt-oss-120b","gpt-oss-120b","Server","51095.2","exact_match: 83.948","1","NVIDIA Grace CPU","2","72","NVIDIA GB200","4","4","./closed/RedHat_Supermicro/results/GB200-NVL4_186GB_Openshift_VLLM/gpt-oss-120b/Server/performance/run_1","vLLM 0.24.0, Red Hat OpenShift 4.21","Red Hat Enterprise Linux CoreOS 9.6","NVIDIA GB200 NVL4 with 2x Grace CPUs (72 ARM cores each), 4x GB200 GPUs (189GB HBM3e each). SMC Lab, Red Hat OpenShift 4.21, RHEL CoreOS 9.6, vLLM 0.24.0","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","llama3_1-8b","llama3.1-8b","Offline","1833.97","ROUGE1: 38.6493  ROUGE2: 15.8388  ROUGEL: 24.4665  ROUGELSUM: 35.6409  GEN_LEN: 8186666","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_128C/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GA-NRT. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","llama3_1-8b","llama3.1-8b","Server","1101.19","ROUGE1: 38.7236  ROUGE2: 15.8895  ROUGEL: 24.4948  ROUGELSUM: 35.7134  GEN_LEN: 8199179","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_128C/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GA-NRT. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_128C","1-node-2S-GNR_128C","whisper","whisper","Offline","2248.15","ACCURACY: 97.89525766917866","1","Intel(R) Xeon(R) 6980P","2","128","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_128C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GA-NRT. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","llama3_1-8b","llama3.1-8b","Offline","1189.89","ROUGE1: 38.7008  ROUGE2: 15.8479  ROUGEL: 24.4355  ROUGELSUM: 35.7034  GEN_LEN: 8193707","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_86C/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GB-NRT. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","llama3_1-8b","llama3.1-8b","Server","516.059","ROUGE1: 38.7059  ROUGE2: 15.8614  ROUGEL: 24.4785  ROUGELSUM: 35.726  GEN_LEN: 8175493","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_86C/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GB-NRT. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","rgat","rgat","Offline","13273.7","acc: 72.520","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_86C/rgat/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GB-NRT. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-2S-GNR_86C","1-node-2S-GNR_86C","whisper","whisper","Offline","1477.4","ACCURACY: 97.89045996917045","1","Intel(R) Xeon(R) 6787P","2","86","N/A","0","0","./closed/Supermicro/results/1-node-2S-GNR_86C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3","SYS-422GB-NRT. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4S-GNR_86C","1-node-4S-GNR_86C","llama3_1-8b","llama3.1-8b","Offline","2063.24","ROUGE1: 38.5949  ROUGE2: 15.7867  ROUGEL: 24.3919  ROUGELSUM: 35.6229  GEN_LEN: 8190845","1","Intel(R) Xeon(R) 6788P","4","86","N/A","0","0","./closed/Supermicro/results/1-node-4S-GNR_86C/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3 LTS","SYS-442B-NR. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4S-GNR_86C","1-node-4S-GNR_86C","llama3_1-8b","llama3.1-8b","Server","966.241","ROUGE1: 38.7724  ROUGE2: 16.0093  ROUGEL: 24.5982  ROUGELSUM: 35.8002  GEN_LEN: 8168484","1","Intel(R) Xeon(R) 6788P","4","86","N/A","0","0","./closed/Supermicro/results/1-node-4S-GNR_86C/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04.3 LTS","SYS-442B-NR. N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4S-GNR_86C","1-node-4S-GNR_86C","whisper","whisper","Offline","2881.28","ACCURACY: 97.9447557970256","1","Intel(R) Xeon(R) 6788P","4","86","N/A","0","0","./closed/Supermicro/results/1-node-4S-GNR_86C/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04.3 LTS","SYS-442B-NR. N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","gpt-oss-120b","gpt-oss-120b","Offline","2063.93","exact_match: 83.780","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/gpt-oss-120b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","MXFP4","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","gpt-oss-120b","gpt-oss-120b","Server","1351.13","exact_match: 84.186","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/gpt-oss-120b/Server/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","MXFP4","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama2-70b-99.9","llama2-70b-99.9","Offline","2484.14","ROUGE1: 44.585  ROUGE2: 22.2008  ROUGEL: 28.8241  TOKENS_PER_SAMPLE: 290.8","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/llama2-70b-99.9/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama2-70b-99.9","llama2-70b-99.9","Server","1759.95","ROUGE1: 44.591  ROUGE2: 22.2044  ROUGEL: 28.8277  TOKENS_PER_SAMPLE: 290.8","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/llama2-70b-99.9/Server/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama3_1-8b","llama3.1-8b","Offline","4596.62","ROUGE1: 38.8447  ROUGE2: 16.0909  ROUGEL: 24.6442  ROUGELSUM: 35.9477  GEN_LEN: 8127563","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/llama3_1-8b/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","llama3_1-8b","llama3.1-8b","Server","4208.73","ROUGE1: 38.8128  ROUGE2: 16.082  ROUGEL: 24.6384  ROUGELSUM: 35.9155  GEN_LEN: 8126189","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/llama3_1-8b/Server/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Tokens/s","INT8","None",""
"Supermicro","available","closed","datacenter","1-node-4x-BMG-B70","1-node-4x-BMG-B70","whisper","whisper","Offline","5854.85","ACCURACY: 97.49980776624375","1","Intel(R) Xeon(R) 6731P","1","32","Intel(R) Arc(TM) Pro B70","4","4","./closed/Supermicro/results/1-node-4x-BMG-B70/whisper/Offline/performance/run_1","PyTorch","Ubuntu 24.04.4 LTS","N/A","closed","0","v6.1","0","False","Samples/s","INT8","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","gpt-oss-120b","gpt-oss-120b","Offline","117005.0","exact_match: 83.113","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/gpt-oss-120b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","gpt-oss-120b","gpt-oss-120b","Server","109184.0","exact_match: 83.861","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/gpt-oss-120b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","llama2-70b-99.9","llama2-70b-99.9","Interactive","73181.5","ROUGE1: 44.507  ROUGE2: 22.1597  ROUGEL: 28.9194  TOKENS_PER_SAMPLE: 284.1","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/llama2-70b-99.9/Interactive/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","1","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","llama2-70b-99.9","llama2-70b-99.9","Offline","102243.0","ROUGE1: 44.5269  ROUGE2: 22.1836  ROUGEL: 28.9207  TOKENS_PER_SAMPLE: 284.3","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/llama2-70b-99.9/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","llama2-70b-99.9","llama2-70b-99.9","Server","103032.0","ROUGE1: 44.5345  ROUGE2: 22.1731  ROUGEL: 28.933  TOKENS_PER_SAMPLE: 284.0","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/llama2-70b-99.9/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","llama3.1-8b","llama3.1-8b","Offline","159060.0","ROUGE1: 38.5437  ROUGE2: 16.0099  ROUGEL: 24.4209  ROUGELSUM: 35.6859  GEN_LEN: 8150015","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/llama3.1-8b/Offline/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Supermicro","available","closed","datacenter","AS-4126GS-NMR-LCC","8xMI355X_2xEPYC-9575F","llama3.1-8b","llama3.1-8b","Server","148710.0","ROUGE1: 38.5048  ROUGE2: 15.9985  ROUGEL: 24.4065  ROUGELSUM: 35.652  GEN_LEN: 8176781","1","AMD EPYC 9575F","2","64","AMD Instinct MI355X 288GB HBM3e","8","8","./closed/Supermicro/results/8xMI355X_2xEPYC-9575F/llama3.1-8b/Server/performance/run_1","vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4","Ubuntu 26.04 LTS","","closed","0","v6.1","0","False","Tokens/s","fp8","None",""
"Supermicro","available","closed","datacenter","AS-8126GS-NB3RT","AS-8126GS-NB3RT_B300-SXM-288GB_TRT","deepseek-r1","deepseek-r1","Offline","69830.0","exact_match: 80.78851412944394  TOKENS_PER_SAMPLE: 3839.5469462169553","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/AS-8126GS-NB3RT_B300-SXM-288GB_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-8126GS-NB3RT","AS-8126GS-NB3RT_B300-SXM-288GB_TRT","deepseek-r1","deepseek-r1","Server","61164.9","exact_match: 81.22151321786691  TOKENS_PER_SAMPLE: 3850.996809480401","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/AS-8126GS-NB3RT_B300-SXM-288GB_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-8126GS-NB3RT","AS-8126GS-NB3RT_B300-SXM-288GB_TRT","gpt-oss-120b","gpt-oss-120b","Offline","108727.0","exact_match: 83.743","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/AS-8126GS-NB3RT_B300-SXM-288GB_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-8126GS-NB3RT","AS-8126GS-NB3RT_B300-SXM-288GB_TRT","gpt-oss-120b","gpt-oss-120b","Server","107288.0","exact_match: 83.354","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/AS-8126GS-NB3RT_B300-SXM-288GB_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-8126GS-NB3RT","AS-8126GS-NB3RT_B300-SXM-288GB_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","110802.0","ROUGE1: 44.8018  ROUGE2: 22.3841  ROUGEL: 29.1588  TOKENS_PER_SAMPLE: 272.7","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/AS-8126GS-NB3RT_B300-SXM-288GB_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","AS-8126GS-NB3RT","AS-8126GS-NB3RT_B300-SXM-288GB_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","106872.0","ROUGE1: 44.828  ROUGE2: 22.4004  ROUGEL: 29.1931  TOKENS_PER_SAMPLE: 272.0","1","AMD EPYC 9575F","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/AS-8126GS-NB3RT_B300-SXM-288GB_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","deepseek-r1","deepseek-r1","Offline","69104.0","exact_match: 81.58614402917047  TOKENS_PER_SAMPLE: 3787.6494986326343","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","deepseek-r1","deepseek-r1","Server","66754.3","exact_match: 80.92525068368278  TOKENS_PER_SAMPLE: 3729.1365086599817","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","gpt-oss-120b","gpt-oss-120b","Offline","107851.0","exact_match: 83.592","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","gpt-oss-120b","gpt-oss-120b","Server","107205.0","exact_match: 83.457","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","llama2-70b-99.9","llama2-70b-99.9","Offline","111247.0","ROUGE1: 44.7966  ROUGE2: 22.3806  ROUGEL: 29.1557  TOKENS_PER_SAMPLE: 272.7","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","llama2-70b-99.9","llama2-70b-99.9","Server","109848.0","ROUGE1: 44.799  ROUGE2: 22.3988  ROUGEL: 29.1884  TOKENS_PER_SAMPLE: 272.3","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","qwen3-vl-235b-a22b","VLM","Offline","128.90612412004543","0.7870632563988416","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/qwen3-vl-235b-a22b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Samples/s","BF16/F8_E4M3/U8","None",""
"Supermicro","available","closed","datacenter","SYS-822GS-NB3RT","SYS-822GS-NB3RT","qwen3-vl-235b-a22b","VLM","Server","111.77383650675426","0.7871852181615411","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B300-SXM-270GB","8","8","./closed/Supermicro/results/SYS-822GS-NB3RT/qwen3-vl-235b-a22b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","1","False","Queries/s","BF16/F8_E4M3/U8","None",""
"TTA","available","closed","datacenter","ManyCoreSoft DeepGadget dg5W (4x NVIDIA RTX PRO 6000 Blackwell 96GB, TensorRT)","DeepGadget_dg5W_RTXPro6000_96GBx4_TRT","llama3.1-8b","llama3.1-8b","Offline","24511.8","ROUGE1: 38.5238  ROUGE2: 15.9869  ROUGEL: 24.3988  ROUGELSUM: 35.6579  GEN_LEN: 8178576","1","AMD EPYC 9124","1","16","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/TTA/results/DeepGadget_dg5W_RTXPro6000_96GBx4_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14.1.48, CUDA 13.1","Rocky Linux 9.7","DeepGadget dg5W; ManyCoreSoft dg5W (SKU2-T1) chassis, GENOAD8X-2T/BCM mainboard; 4x NVIDIA RTX PRO 6000 Blackwell Server Edition (TGP 600W each); direct liquid cooling (DLC)","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"TTA","available","closed","datacenter","ManyCoreSoft DeepGadget dg5W (4x NVIDIA RTX PRO 6000 Blackwell 96GB, TensorRT)","DeepGadget_dg5W_RTXPro6000_96GBx4_TRT","llama3.1-8b","llama3.1-8b","Server","24510.5","ROUGE1: 38.5242  ROUGE2: 15.9887  ROUGEL: 24.3995  ROUGELSUM: 35.6582  GEN_LEN: 8178682","1","AMD EPYC 9124","1","16","NVIDIA RTX PRO 6000 Blackwell Server Edition","4","4","./closed/TTA/results/DeepGadget_dg5W_RTXPro6000_96GBx4_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14.1.48, CUDA 13.1","Rocky Linux 9.7","DeepGadget dg5W; ManyCoreSoft dg5W (SKU2-T1) chassis, GENOAD8X-2T/BCM mainboard; 4x NVIDIA RTX PRO 6000 Blackwell Server Edition (TGP 600W each); direct liquid cooling (DLC)","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","gpt-oss-120b","gpt-oss-120b","Offline","14130.6","exact_match: 83.235","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","gpt-oss-120b","gpt-oss-120b","Server","13854.0","exact_match: 83.289","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","llama2-70b-99.9","llama2-70b-99.9","Offline","14636.8","ROUGE1: 44.781  ROUGE2: 22.3679  ROUGEL: 29.1539  TOKENS_PER_SAMPLE: 273.1","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/llama2-70b-99.9/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","llama2-70b-99.9","llama2-70b-99.9","Server","14495.6","ROUGE1: 44.7255  ROUGE2: 22.327  ROUGEL: 29.1172  TOKENS_PER_SAMPLE: 274.2","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/llama2-70b-99.9/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","llama3.1-8b","llama3.1-8b","Offline","21266.1","ROUGE1: 38.5346  ROUGE2: 15.7631  ROUGEL: 24.3218  ROUGELSUM: 35.5543  GEN_LEN: 8191123","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","llama3.1-8b","llama3.1-8b","Server","20420.4","ROUGE1: 38.5378  ROUGE2: 15.764  ROUGEL: 24.3237  ROUGELSUM: 35.5571  GEN_LEN: 8191033","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","B300-SXM-270GBx1_TRT","B300-SXM-270GBx1_TRT","whisper","whisper","Offline","7058.95","ACCURACY: 97.85680976807484","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B300-SXM-270GB","1","1","./closed/VibeHPC/results/B300-SXM-270GBx1_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Offline","36639.6","exact_match: 81.22151321786691  TOKENS_PER_SAMPLE: 3770.8865086599817","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","deepseek-r1","deepseek-r1","Server","33415.2","exact_match: 81.83682771194167  TOKENS_PER_SAMPLE: 3750.4530537830447","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","llama3.1-8b","llama3.1-8b","Offline","161767.0","ROUGE1: 38.5395  ROUGE2: 15.7648  ROUGEL: 24.3262  ROUGELSUM: 35.5597  GEN_LEN: 8190944","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/llama3.1-8b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","llama3.1-8b","llama3.1-8b","Server","156867.0","ROUGE1: 38.5355  ROUGE2: 15.7593  ROUGEL: 24.3212  ROUGELSUM: 35.5549  GEN_LEN: 8191008","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/llama3.1-8b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","Offline","0.0783824","vbench_score: 70.1187","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Samples/s","fp32","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","wan-2.2-t2v-a14b","text_to_video","SingleStream","21.279754716","vbench_score: 70.3113","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/wan-2.2-t2v-a14b/SingleStream/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Latency (s)","fp32","None",""
"VibeHPC","available","closed","datacenter","NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","B300-SXM-270GBx8_TRT","whisper","whisper","Offline","53267.2","ACCURACY: 97.85680976807484","1","Intel(R) Xeon(R) 6776P","2","48","NVIDIA B300-SXM-270GB","8","8","./closed/VibeHPC/results/B300-SXM-270GBx8_TRT/whisper/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1","Ubuntu 24.04","B300 TGP 1100W","closed","0","v6.1","0","False","Samples/s","fp16","None",""
"Wiwynn","available","closed","datacenter","OneW GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","gpt-oss-120b","gpt-oss-120b","Offline","63990.5","exact_match: 83.440","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Wiwynn/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/gpt-oss-120b/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"Wiwynn","available","closed","datacenter","OneW GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","GB300-NVL72_GB300-288GB_aarch64x4_TRT","gpt-oss-120b","gpt-oss-120b","Server","60466.2","exact_match: 83.187","1","NVIDIA Grace CPU","2","144","NVIDIA GB300","4","4","./closed/Wiwynn/results/GB300-NVL72_GB300-288GB_aarch64x4_TRT/gpt-oss-120b/Server/performance/run_1","TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf","Ubuntu 24.04","NVIDIA GB300 NVL72, 1400W TGP Per GPU","closed","0","v6.1","0","False","Tokens/s","fp4","None",""
"AMD","available","open","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC_9575F","dlrm-v3","dlrm-v3","Offline","16363.0","DLRM_NE: 0.8678795984098008  DLRM_ACC: 0.6962884623080815  DLRM_AUC: 0.7861666437936649","1","AMD EPYC 9575F","2","64 cores per socket, 128 total cores, 256 threads","AMD Instinct MI355X","8","8","./open/AMD/results/8xMI355X_2xEPYC_9575F/dlrm-v3/Offline/performance/run_1","PyTorch 2.10.0+rocm7.2.3.git1a270074, Triton 3.6.0, ROCm 7.2.3","Ubuntu 24.04.3 LTS (Noble Numbat), kernel 6.8.0-134-generic","Measured on current chi2810-class host reporting hostname chi2878: 8x MI355X / 2x EPYC 9575F, 256 online CPUs, SMT enabled, CPU governor performance. Known-good host KMD floor: amdgpu 6.16.6 or newer.. DLRM-v3 ROCm C1-on stack: HSTU sliding-window attention (window 1024), batch 64, degree-5 SiLU gate, uniform-target metadata, and FP8 quantization. MLPerf Inference v6.1 seeds with LoadGen 6.0.17. Server target 16,200 q/s; Offline target 17,000 q/s.","open","0","v6.1","0","False","Samples/s","fp16,bf16,fp8","None",""
"AMD","available","open","datacenter","8xMI355X_2xEPYC_9575F","8xMI355X_2xEPYC_9575F","dlrm-v3","dlrm-v3","Server","16194.4","DLRM_NE: 0.8678772746839061  DLRM_ACC: 0.6962803331970739  DLRM_AUC: 0.7861629372250531","1","AMD EPYC 9575F","2","64 cores per socket, 128 total cores, 256 threads","AMD Instinct MI355X","8","8","./open/AMD/results/8xMI355X_2xEPYC_9575F/dlrm-v3/Server/performance/run_1","PyTorch 2.10.0+rocm7.2.3.git1a270074, Triton 3.6.0, ROCm 7.2.3","Ubuntu 24.04.3 LTS (Noble Numbat), kernel 6.8.0-134-generic","Measured on current chi2810-class host reporting hostname chi2878: 8x MI355X / 2x EPYC 9575F, 256 online CPUs, SMT enabled, CPU governor performance. Known-good host KMD floor: amdgpu 6.16.6 or newer.. DLRM-v3 ROCm C1-on stack: HSTU sliding-window attention (window 1024), batch 64, degree-5 SiLU gate, uniform-target metadata, and FP8 quantization. MLPerf Inference v6.1 seeds with LoadGen 6.0.17. Server target 16,200 q/s; Offline target 17,000 q/s.","open","0","v6.1","0","False","Queries/s","fp16,bf16,fp8","None",""
"HPE","available","open","datacenter","NVIDIA GB200 NVL72 by HPE (16x GB200-186GB_aarch64, TensorRT)","HPE_GB200-NVL72_GB200-186GB_aarch64x16_TRT","deepseek-r1","deepseek-r1","Offline","126223.0","exact_match: 81.38103919781221  TOKENS_PER_SAMPLE: 3823.224020054695","4","NVIDIA Grace CPU (Arm Neoverse-V2)","2","144","NVIDIA GB200","4","16","./open/HPE/results/HPE_GB200-NVL72_GB200-186GB_aarch64x16_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k","NVIDIA GB200 NVL72 platform by HPE; 1200 W GPU power limit. Two dep8 replicas across four nodes; CentML DeepSeek-R1 NVFP4 v2 checkpoint","open","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","open","datacenter","NVIDIA GB200 NVL72 by HPE (32x GB200-186GB_aarch64, TensorRT)","HPE_GB200-NVL72_GB200-186GB_aarch64x32_TRT","deepseek-r1","deepseek-r1","Offline","245352.0","exact_match: 81.8824065633546  TOKENS_PER_SAMPLE: 3798.4737921604374","8","NVIDIA Grace CPU (Arm Neoverse-V2)","2","144","NVIDIA GB200","4","32","./open/HPE/results/HPE_GB200-NVL72_GB200-186GB_aarch64x32_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k","NVIDIA GB200 NVL72 platform by HPE; 1200 W GPU power limit. Four dep8 replicas across eight nodes; CentML DeepSeek-R1 NVFP4 v2 checkpoint","open","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","open","datacenter","NVIDIA GB200 NVL72 by HPE (56x GB200-186GB_aarch64, TensorRT)","HPE_GB200-NVL72_GB200-186GB_aarch64x56_TRT","deepseek-r1","deepseek-r1","Offline","336125.0","exact_match: 80.9936189608022  TOKENS_PER_SAMPLE: 3836.976071103008","14","NVIDIA Grace CPU (Arm Neoverse-V2)","2","144","NVIDIA GB200","4","56","./open/HPE/results/HPE_GB200-NVL72_GB200-186GB_aarch64x56_TRT/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0","Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k","NVIDIA GB200 NVL72 platform by HPE; 1200 W GPU power limit. The LoadGen task shares one inference node CPU allocation and is not an additional SUT node.. Seven dep8 replicas across fourteen inference nodes; CentML DeepSeek-R1 NVFP4 v2 checkpoint","open","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","open","edge","HPE ProLiant DL345 Gen12 (4x NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB, TensorRT)","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_TRT_open","llama3.1-8b-edge","llama3.1-8b-edge","Offline","9557.69","ROUGE1: 38.7947  ROUGE2: 16.0822  ROUGEL: 24.5817  ROUGELSUM: 35.7134  GEN_LEN: 8173143","1","AMD EPYC 9375F 32-Core Processor","1","32","NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB","4","4","./open/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_TRT_open/llama3.1-8b-edge/Offline/performance/run_1","TensorRT 10.15.1.29, CUDA 13.2","Ubuntu 24.04.4","","open","0","v6.1","0","False","Tokens/s","fp4","None",""
"HPE","available","open","edge","HPE ProLiant DL345 Gen12 (4x NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB, TensorRT)","HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_TRT_open","llama3.1-8b-edge","llama3.1-8b-edge","SingleStream","1203.785699","ROUGE1: 38.7883  ROUGE2: 16.056  ROUGEL: 24.5055  ROUGELSUM: 35.724  GEN_LEN: 8183134","1","AMD EPYC 9375F 32-Core Processor","1","32","NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB","4","4","./open/HPE/results/HPE_PROLIANT_DL345_GEN12_RTX_PRO_4500_PCIE_32GBx4_TRT_open/llama3.1-8b-edge/SingleStream/performance/run_1","TensorRT 10.15.1.29, CUDA 13.2","Ubuntu 24.04.4","","open","0","v6.1","0","False","Latency (ms)","fp4","None",""
"HPE","available","open","edge","HPE ProLiant Compute DL380 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB)","HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT_open","resnet50","resnet","multistream","5.030642","acc: 77.805","1","Intel(R) Xeon(R) 6762P","2","64","NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB","1","1","./open/HPE/results/HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT_open/resnet50/multistream/performance/run_1","ONNX Runtime v1.28.0","Ubuntu 24.04.3","","open","0","v6.1","1","False","Latency (ms)","fp32","None",""
"HPE","available","open","edge","HPE ProLiant Compute DL380 Gen12 (1x NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB)","HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT_open","resnet50","resnet","singlestream","1.480867","acc: 77.536","1","Intel(R) Xeon(R) 6762P","2","64","NVIDIA RTX PRO 4500 Blackwell-PCIe-32GB","1","1","./open/HPE/results/HPE_PROLIANT_DL380_RTX_PRO_4500_PCIE_32GBx1_TRT_open/resnet50/singlestream/performance/run_1","ONNX Runtime v1.28.0","Ubuntu 24.04.3","","open","0","v6.1","1","False","Latency (ms)","fp32","None",""
"HPE","available","open","datacenter","HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","HPE_XD690_B300_SXM_270GBx8_TRT_open","deepseek-r1","deepseek-r1","Offline","71053.4","exact_match: 81.08477666362808  TOKENS_PER_SAMPLE: 3818.841841385597","1","Intel(R) Xeon(R) 6776P","2","64","NVIDIA B300-SXM-270GB","8","8","./open/HPE/results/HPE_XD690_B300_SXM_270GBx8_TRT_open/deepseek-r1/Offline/performance/run_1","TensorRT 10.14, CUDA 13.2","RHEL 9.4","B300 TGP 1100W","open","0","v6.1","0","False","Tokens/s","fp4","None",""
"Krai","available","open","datacenter","HPE Cray XD670 (8x H200-SXM-141GB, Krai/Gentic)","h200","qwen3-vl-235b-a22b","VLM","Offline","9.221680429087034","0.7878368975913989","1","Intel(R) Xeon(R) Platinum 8562Y+","2","32","NVIDIA H200-SXM-141GB","8","8","./open/Krai/results/h200/qwen3-vl-235b-a22b/Offline/performance/run_1","Krai/Gentic","Red Hat Enterprise Linux release 9.4 (Plow) (Linux kernel: 5.14.0-427.13.1.el9_4.x86_64 #1 SMP PREEMPT_DYNAMIC Wed Apr 10 10:29:16 EDT 2024 x86_64 x86_64 x86_64 GNU/Linux)","Powered by Krai/Gentic, next-gen automation","open","0","v6.1","1","False","Samples/s","bf16","None",""
"Lambda","available","open","edge","Lambda 8x B200-SXM-180GB (SGLang, Kimi-K2.6 edge-agentic / BFCL v4)","B200-SXM-180GBx8_SGLANG","qwen3.6-27b","agentic edge","SingleStream","770.827765960278","0.8683417085427135","1","Intel(R) Xeon(R) Platinum 8570","2","56","NVIDIA B200-SXM-180GB","8","8","./open/Lambda/results/B200-SXM-180GBx8_SGLANG/qwen3.6-27b/SingleStream/performance/run_1","SGLang 0.5.15.post1, CUDA 13.0","Ubuntu 24.04.4 LTS","8x B200-SXM-180GB (Blackwell). OPEN division. Model = moonshotai/Kimi-K2.6 (kimi_k25, fp8), served with SGLang TP=8 on 8x B200, reasoning disabled (chat_template thinking=false, the reference 'reasoning off' analog). Workload = the MLPerf edge-agentic reference (mlcommons/endpoints example 11_Edge_Agentic_Example): BFCL v4 single-turn function-calling accuracy + agentic-coding inline-checker performance replay, all reference invariants unchanged (temperature 0, max_new_tokens 1024, single-stream). Benchmark labeled kimi-k2.6-edge-agentic. All metrics are Kimi-K2.6's real measured numbers.","open","0","v6.1","1","False","Latency (ms)","fp8","None",""
"Orrick Industries LLC","rdi","open","datacenter","Orrick Optimized 6x NVIDIA B200","b200_6x","llama3.1-8b","llama3.1-8b","Offline","274927.0","ROUGE1: 36.869  ROUGE2: 14.1535  ROUGEL: 23.5033  ROUGELSUM: 33.7502","1","INTEL(R) XEON(R) PLATINUM 8568Y+","2","48","NVIDIA B200-SXM-180GB","6","6","./open/Orrick Industries LLC/results/b200_6x/llama3.1-8b/Offline/performance/run_1","TensorRT-LLM 1.3.0rc20","Ubuntu 24.04.3 LTS (container)","6x NVIDIA B200, rented cloud instance. Orrick proprietary inference optimization. NVFP4 weights with FP8 KV cache. Implementation details are proprietary and not disclosed.","open","0","v6.1","0","False","Tokens/s","nvfp4","None",""
"Orrick Industries LLC","rdi","open","datacenter","Orrick Optimized 6x NVIDIA B200 L2","b200_l2_6x","llama2-70b-99.9","llama2-70b-99.9","Offline","130928.0","ROUGE1: 41.797  ROUGE2: 19.7311  ROUGEL: 26.6864  TOKENS_PER_SAMPLE: 281.9","1","INTEL(R) XEON(R) PLATINUM 8568Y+","2","48","NVIDIA B200-SXM-180GB","6","6","./open/Orrick Industries LLC/results/b200_l2_6x/llama2-70b-99.9/Offline/performance/run_1","TensorRT-LLM 1.3.0rc20","Ubuntu 24.04.3 LTS (container)","6x NVIDIA B200, rented cloud instance. Orrick proprietary inference optimization. NVFP4 weights with FP8 KV cache. Implementation details are proprietary and not disclosed.","open","0","v6.1","0","False","Tokens/s","nvfp4","None",""
"ScitiX","available","open","datacenter","ScitiX B200 (8x B200-SXM-180GB, TensorRT-LLM)","B200-SXM-180GBx8_TRT","deepseek-r1","deepseek-r1","Offline","60410.9","exact_match: 81.70009115770283  TOKENS_PER_SAMPLE: 3727.5230173199634","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B200-SXM-180GB","8","8","./open/ScitiX/results/B200-SXM-180GBx8_TRT/deepseek-r1/Offline/performance/run_1","TensorRT-LLM 1.3.0rc0, CUDA 13.1","Ubuntu 24.04.3 LTS","Closed division; DeepSeek-R1 NVFP4 weights + FP8 KV; aggregated 8-GPU TP=8 with MTP (num_nextn_predict_layers=1) speculative decoding","open","0","v6.1","0","False","Tokens/s","fp4","None",""
"ScitiX","available","open","datacenter","ScitiX B200 (8x B200-SXM-180GB, TensorRT-LLM)","B200-SXM-180GBx8_TRT","deepseek-r1","deepseek-r1","Server","59668.3","exact_match: 81.17593436645396  TOKENS_PER_SAMPLE: 3739.3393345487693","1","Intel(R) Xeon(R) 6767P","2","64","NVIDIA B200-SXM-180GB","8","8","./open/ScitiX/results/B200-SXM-180GBx8_TRT/deepseek-r1/Server/performance/run_1","TensorRT-LLM 1.3.0rc0, CUDA 13.1","Ubuntu 24.04.3 LTS","Closed division; DeepSeek-R1 NVFP4 weights + FP8 KV; aggregated 8-GPU TP=8 with MTP (num_nextn_predict_layers=1) speculative decoding","open","0","v6.1","0","False","Tokens/s","fp4","None",""