{
  "assumptions": {
    "hours_per_month": 730,
    "hosted": "always-on rented GPU; throughput at the most concurrent users that meet the latency SLO (p99 TTFT and median inter-token latency)",
    "api": "list price per token for the same model on each provider (input and output priced separately)"
  },
  "gpu_offers": [
    {
      "provider": "runpod",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 2.69,
      "min_gpus": 1,
      "product": "H100 SXM, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 3.49,
      "min_gpus": 1,
      "product": "H100 SXM, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "vast",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 2.4024,
      "min_gpus": 1,
      "product": "H100 SXM, marketplace median",
      "source": "https://console.vast.ai/api/v0/bundles/",
      "checked_at": "2026-10-19",
      "notes": "median of 4 verified rentable 1x offers; marketplace prices move daily"
    },
    {
      "provider": "nebius",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 3.85,
      "min_gpus": 1,
      "product": "HGX H100",
      "source": "https://nebius.com/prices",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 4.29,
      "min_gpus": 1,
      "product": "1x H100 SXM",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 3.99,
      "min_gpus": 8,
      "product": "8x H100 SXM",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "gcp",
      "gpu_type": "H100-80GB",
      "provisioning": "spot",
      "usd_per_gpu_hour": 6.571,
      "min_gpus": 1,
      "product": "a3-highgpu-1g Spot, us-central1",
      "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
      "checked_at": "2026-09-26",
      "notes": "GPU $5.8632 + 26 vCPU + 234 GiB from the Cloud Billing catalog; boot disk excluded"
    },
    {
      "provider": "gcp",
      "gpu_type": "H100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 10.9791,
      "min_gpus": 1,
      "product": "a3-highgpu-1g, us-central1",
      "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
      "checked_at": "2026-09-26",
      "notes": "1g/2g/4g shapes are Spot/Flex-start only; on-demand-equivalent list price"
    },
    {
      "provider": "runpod",
      "gpu_type": "H100-PCIe-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.99,
      "min_gpus": 1,
      "product": "H100 PCIe, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "H100-PCIe-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 2.89,
      "min_gpus": 1,
      "product": "H100 PCIe, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "H100-PCIe-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 3.29,
      "min_gpus": 1,
      "product": "1x H100 PCIe",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "H100-NVL-94GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 2.59,
      "min_gpus": 1,
      "product": "H100 NVL, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "H100-NVL-94GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 3.19,
      "min_gpus": 1,
      "product": "H100 NVL, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "H200-141GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 3.59,
      "min_gpus": 1,
      "product": "H200, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "H200-141GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 4.59,
      "min_gpus": 1,
      "product": "H200, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "nebius",
      "gpu_type": "H200-141GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 4.5,
      "min_gpus": 1,
      "product": "HGX H200",
      "source": "https://nebius.com/prices",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "B200-180GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 5.98,
      "min_gpus": 1,
      "product": "B200, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "B200-180GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 6.79,
      "min_gpus": 1,
      "product": "B200, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "B200-180GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 6.99,
      "min_gpus": 1,
      "product": "1x B200 SXM6",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "nebius",
      "gpu_type": "B200-180GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 7.15,
      "min_gpus": 1,
      "product": "HGX B200",
      "source": "https://nebius.com/prices",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "GH200-96GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 2.29,
      "min_gpus": 1,
      "product": "1x GH200",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "A100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.39,
      "min_gpus": 1,
      "product": "A100 SXM, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "A100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.59,
      "min_gpus": 1,
      "product": "A100 SXM, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "A100-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 2.79,
      "min_gpus": 8,
      "product": "8x A100 SXM 80GB",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "A100-PCIe-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.19,
      "min_gpus": 1,
      "product": "A100 PCIe, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "A100-PCIe-80GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.59,
      "min_gpus": 1,
      "product": "A100 PCIe, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "A100-40GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.99,
      "min_gpus": 1,
      "product": "1x A100 SXM 40GB",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "L40S-48GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 0.79,
      "min_gpus": 1,
      "product": "L40S, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "L40S-48GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.09,
      "min_gpus": 1,
      "product": "L40S, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "nebius",
      "gpu_type": "L40S-48GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.55,
      "min_gpus": 1,
      "product": "L40S (Intel)",
      "source": "https://nebius.com/prices",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "lambda",
      "gpu_type": "A10-24GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 1.29,
      "min_gpus": 1,
      "product": "1x A10",
      "source": "https://lambda.ai/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "RTX4090-24GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 0.34,
      "min_gpus": 1,
      "product": "RTX 4090, Community Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    },
    {
      "provider": "runpod",
      "gpu_type": "RTX4090-24GB",
      "provisioning": "on-demand",
      "usd_per_gpu_hour": 0.74,
      "min_gpus": 1,
      "product": "RTX 4090, Secure Cloud",
      "source": "https://www.runpod.io/pricing",
      "checked_at": "2026-09-25",
      "notes": null
    }
  ],
  "api_offers": [
    {
      "model": "Qwen/Qwen3-14B",
      "provider": "Fireworks",
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "quantization": null,
      "context_length": null,
      "source": "https://docs.fireworks.ai/serverless/pricing",
      "fetched_at": "2026-09-25",
      "notes": "4B-16B size-tier price (Standard tier); serverless availability varies"
    },
    {
      "model": "Qwen/Qwen3-14B",
      "provider": "NextBit",
      "usd_per_1m_input": 0.1,
      "usd_per_1m_output": 0.22,
      "quantization": "int4",
      "context_length": 40960,
      "source": "https://openrouter.ai/qwen/qwen3-14b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3-14B",
      "provider": "DeepInfra",
      "usd_per_1m_input": 0.12,
      "usd_per_1m_output": 0.24,
      "quantization": "fp8",
      "context_length": 40960,
      "source": "https://openrouter.ai/qwen/qwen3-14b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3-8B",
      "provider": "Fireworks",
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "quantization": null,
      "context_length": null,
      "source": "https://docs.fireworks.ai/serverless/pricing",
      "fetched_at": "2026-09-25",
      "notes": "4B-16B size-tier price (Standard tier)"
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Tenstorrent",
      "usd_per_1m_input": 0.02,
      "usd_per_1m_output": 0.5,
      "quantization": "fp4",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Chutes",
      "usd_per_1m_input": 0.039,
      "usd_per_1m_output": 1.25,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Near AI",
      "usd_per_1m_input": 0.04,
      "usd_per_1m_output": 1.35,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Cerebras",
      "usd_per_1m_input": 0.99,
      "usd_per_1m_output": 1.49,
      "quantization": "fp16",
      "context_length": 65536,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "DeepInfra",
      "usd_per_1m_input": 0.15,
      "usd_per_1m_output": 1.875,
      "quantization": "bf16",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Ionstream",
      "usd_per_1m_input": 0.04,
      "usd_per_1m_output": 1.95,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "AkashML",
      "usd_per_1m_input": 0.225,
      "usd_per_1m_output": 1.98,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Darkbloom",
      "usd_per_1m_input": 0.035,
      "usd_per_1m_output": 2.2,
      "quantization": "fp4",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Parasail",
      "usd_per_1m_input": 0.24,
      "usd_per_1m_output": 2.2,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Reka",
      "usd_per_1m_input": 0.04,
      "usd_per_1m_output": 2.2,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Wafer",
      "usd_per_1m_input": 0.04,
      "usd_per_1m_output": 2.3,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "DekaLLM",
      "usd_per_1m_input": 0.04,
      "usd_per_1m_output": 2.5,
      "quantization": null,
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Mancer 2",
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 2.5,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "ModelRun",
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 2.5,
      "quantization": "fp4",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Phala",
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 2.5,
      "quantization": null,
      "context_length": 1000000,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Alibaba",
      "usd_per_1m_input": 0.425,
      "usd_per_1m_output": 2.55,
      "quantization": null,
      "context_length": 1000000,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "CoreWeave",
      "usd_per_1m_input": 0.4,
      "usd_per_1m_output": 3.0,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Novita",
      "usd_per_1m_input": 0.42,
      "usd_per_1m_output": 3.0,
      "quantization": null,
      "context_length": 1000000,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Cloudflare",
      "usd_per_1m_input": 0.45,
      "usd_per_1m_output": 3.2,
      "quantization": null,
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    },
    {
      "model": "Qwen/Qwen3.8-27B",
      "provider": "Venice",
      "usd_per_1m_input": 0.45,
      "usd_per_1m_output": 3.2,
      "quantization": "fp8",
      "context_length": 262144,
      "source": "https://openrouter.ai/qwen/qwen3.8-27b/providers",
      "fetched_at": "2026-10-19",
      "notes": null
    }
  ],
  "deployments": [
    {
      "run_id": "20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "fp8",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "prefill-1024-128",
      "input_len": 1024,
      "output_len": 128,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 108,
      "max_users_kv_cache": 373,
      "capacity_is_lower_bound": false,
      "requests_per_s": 32.25802607049752,
      "output_tokens_per_s": 4129.027337023683,
      "total_tokens_per_s": 37161.246033213145,
      "ttft_p99_ms": 1897.8023735601664,
      "itl_median_ms": 11.766713999804779,
      "accuracy": {
        "mmlu_pro": 0.6263297872340425,
        "gpqa_diamond_cot_zeroshot": 0.45454545454545453,
        "gsm8k": 0.8642911296436695,
        "minerva_math500": 0.822,
        "ifeval": 0.8354898336414048,
        "arc_challenge_chat": 0.8762798634812287
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.16161998428771984,
          "usd_per_1m_total_tokens": 0.017957776031968874,
          "usd_per_1k_requests": 0.02068735798882814
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.18096809762486116,
          "usd_per_1m_total_tokens": 0.020107566402762348,
          "usd_per_1k_requests": 0.023163916495982227
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.23478760621218045,
          "usd_per_1m_total_tokens": 0.026087511801353384,
          "usd_per_1k_requests": 0.030052813595159097
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.2590063850764741,
          "usd_per_1m_total_tokens": 0.02877848723071935,
          "usd_per_1k_requests": 0.03315281728978869
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.28860711479949974,
          "usd_per_1m_total_tokens": 0.03206745719994441,
          "usd_per_1k_requests": 0.036941710694335966
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.44199271427335973,
          "usd_per_1m_total_tokens": 0.04911030158592886,
          "usd_per_1k_requests": 0.05657506742699004
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.44205998865909385,
          "usd_per_1m_total_tokens": 0.0491177765176771,
          "usd_per_1k_requests": 0.05658367854836401
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 0.7386122084137967,
          "usd_per_1m_total_tokens": 0.08206802315708853,
          "usd_per_1k_requests": 0.09454236267696597
        }
      ]
    },
    {
      "run_id": "20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "fp8",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "decode-128-1024",
      "input_len": 128,
      "output_len": 1024,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 472,
      "max_users_kv_cache": 373,
      "capacity_is_lower_bound": false,
      "requests_per_s": 14.74949037164476,
      "output_tokens_per_s": 15103.478140564233,
      "total_tokens_per_s": 16991.412908134764,
      "ttft_p99_ms": 1886.467315499949,
      "itl_median_ms": 27.4143289998392,
      "accuracy": {
        "mmlu_pro": 0.6263297872340425,
        "gpqa_diamond_cot_zeroshot": 0.45454545454545453,
        "gsm8k": 0.8642911296436695,
        "minerva_math500": 0.822,
        "ifeval": 0.8354898336414048,
        "arc_challenge_chat": 0.8762798634812287
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.04418408310474128,
          "usd_per_1m_total_tokens": 0.0392747405375478,
          "usd_per_1k_requests": 0.04524450109925507
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.049473519626937246,
          "usd_per_1m_total_tokens": 0.04397646189061088,
          "usd_per_1k_requests": 0.050660884097983735
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.06418683401413047,
          "usd_per_1m_total_tokens": 0.05705496356811598,
          "usd_per_1k_requests": 0.0657273180304696
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.07080782548836743,
          "usd_per_1m_total_tokens": 0.06294028932299327,
          "usd_per_1k_requests": 0.07250721330008825
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.0789001484013237,
          "usd_per_1m_total_tokens": 0.07013346524562107,
          "usd_per_1k_requests": 0.08079375196295546
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.12083309440482443,
          "usd_per_1m_total_tokens": 0.1074071950265106,
          "usd_per_1k_requests": 0.12373308867054021
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.1208514860478084,
          "usd_per_1m_total_tokens": 0.10742354315360746,
          "usd_per_1k_requests": 0.1237519217129558
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 0.20192368748554154,
          "usd_per_1m_total_tokens": 0.17948772220937026,
          "usd_per_1k_requests": 0.20676985598519454
        }
      ]
    },
    {
      "run_id": "20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "fp8",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "balanced-2048-2048",
      "input_len": 2048,
      "output_len": 2048,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 52,
      "max_users_kv_cache": 105,
      "capacity_is_lower_bound": false,
      "requests_per_s": 1.7837398767000137,
      "output_tokens_per_s": 3653.099267481628,
      "total_tokens_per_s": 7306.198534963256,
      "ttft_p99_ms": 1928.2087817799038,
      "itl_median_ms": 13.32839399947261,
      "accuracy": {
        "mmlu_pro": 0.6263297872340425,
        "gpqa_diamond_cot_zeroshot": 0.45454545454545453,
        "gsm8k": 0.8642911296436695,
        "minerva_math500": 0.822,
        "ifeval": 0.8354898336414048,
        "arc_challenge_chat": 0.8762798634812287
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.18267593746319938,
          "usd_per_1m_total_tokens": 0.09133796873159969,
          "usd_per_1k_requests": 0.37412031992463235
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.20454473517149782,
          "usd_per_1m_total_tokens": 0.10227236758574891,
          "usd_per_1k_requests": 0.4189076176312275
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.26537588317789124,
          "usd_per_1m_total_tokens": 0.13268794158894562,
          "usd_per_1k_requests": 0.5434898087483212
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.29274989978076826,
          "usd_per_1m_total_tokens": 0.14637494989038413,
          "usd_per_1k_requests": 0.5995517947510134
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.3262070311842846,
          "usd_per_1m_total_tokens": 0.1631035155921423,
          "usd_per_1k_requests": 0.6680719998654149
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.4995758030025058,
          "usd_per_1m_total_tokens": 0.2497879015012529,
          "usd_per_1k_requests": 1.0231312445491318
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.4996518419375137,
          "usd_per_1m_total_tokens": 0.24982592096875686,
          "usd_per_1k_requests": 1.023286972288028
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 0.8348390713462422,
          "usd_per_1m_total_tokens": 0.4174195356731211,
          "usd_per_1k_requests": 1.7097504181171042
        }
      ]
    },
    {
      "run_id": "20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "fp8",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "longctx-8192-512",
      "input_len": 8192,
      "output_len": 512,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 10,
      "max_users_kv_cache": 49,
      "capacity_is_lower_bound": false,
      "requests_per_s": 1.5368731646298843,
      "output_tokens_per_s": 786.8790602905008,
      "total_tokens_per_s": 13376.944024938513,
      "ttft_p99_ms": 1709.6512999506376,
      "itl_median_ms": 9.506381000392139,
      "accuracy": {
        "mmlu_pro": 0.6263297872340425,
        "gpqa_diamond_cot_zeroshot": 0.45454545454545453,
        "gsm8k": 0.8642911296436695,
        "minerva_math500": 0.822,
        "ifeval": 0.8354898336414048,
        "arc_challenge_chat": 0.8762798634812287
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.8480761110696815,
          "usd_per_1m_total_tokens": 0.04988683006292244,
          "usd_per_1k_requests": 0.43421496886767696
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.9496023721184829,
          "usd_per_1m_total_tokens": 0.055858963065793105,
          "usd_per_1k_requests": 0.4861964145246632
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 1.2320119995143144,
          "usd_per_1m_total_tokens": 0.07247129408907732,
          "usd_per_1k_requests": 0.630790143751329
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 1.3590963318424385,
          "usd_per_1m_total_tokens": 0.0799468430495552,
          "usd_per_1k_requests": 0.6958573219033285
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 1.5144216269101456,
          "usd_per_1m_total_tokens": 0.0890836251123615,
          "usd_per_1k_requests": 0.7753838729779945
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 2.319289064988265,
          "usd_per_1m_total_tokens": 0.13642876852872146,
          "usd_per_1k_requests": 1.1874760012739916
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 2.3196420770225092,
          "usd_per_1m_total_tokens": 0.13644953394250056,
          "usd_per_1k_requests": 1.1876567434355247
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 3.875754425176965,
          "usd_per_1m_total_tokens": 0.2279855544221744,
          "usd_per_1k_requests": 1.9843862656906062
        }
      ]
    },
    {
      "run_id": "20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "fp8",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "chat-128-128",
      "input_len": 128,
      "output_len": 128,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 512,
      "max_users_kv_cache": 1680,
      "capacity_is_lower_bound": true,
      "requests_per_s": 129.23401451464562,
      "output_tokens_per_s": 16541.95385787464,
      "total_tokens_per_s": 33083.90771574928,
      "ttft_p99_ms": 1506.4633105300436,
      "itl_median_ms": 19.674075499949595,
      "accuracy": {
        "mmlu_pro": 0.6263297872340425,
        "gpqa_diamond_cot_zeroshot": 0.45454545454545453,
        "gsm8k": 0.8642911296436695,
        "minerva_math500": 0.822,
        "ifeval": 0.8354898336414048,
        "arc_challenge_chat": 0.8762798634812287
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.040341868866697134,
          "usd_per_1m_total_tokens": 0.020170934433348567,
          "usd_per_1k_requests": 0.0051637592149372325
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.04517134001474162,
          "usd_per_1m_total_tokens": 0.02258567000737081,
          "usd_per_1k_requests": 0.0057819315218869275
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.058605195781207535,
          "usd_per_1m_total_tokens": 0.029302597890603767,
          "usd_per_1k_requests": 0.007501465059994565
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.06465043087611719,
          "usd_per_1m_total_tokens": 0.032325215438058594,
          "usd_per_1k_requests": 0.008275255152143
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.07203905154767344,
          "usd_per_1m_total_tokens": 0.03601952577383672,
          "usd_per_1k_requests": 0.0092209985981022
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T225750Z_h100-80gbx1_qwen3-8b_fp8",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.11032554048210129,
          "usd_per_1m_total_tokens": 0.055162770241050645,
          "usd_per_1k_requests": 0.014121669181708965
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.11034233280180936,
          "usd_per_1m_total_tokens": 0.05517116640090468,
          "usd_per_1k_requests": 0.014123818598631597
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 0.18436455730700735,
          "usd_per_1m_total_tokens": 0.09218227865350367,
          "usd_per_1k_requests": 0.023598663335296942
        }
      ]
    },
    {
      "run_id": "20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "bf16",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "chat-128-128",
      "input_len": 128,
      "output_len": 128,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 512,
      "max_users_kv_cache": 1472,
      "capacity_is_lower_bound": true,
      "requests_per_s": 113.59005371064828,
      "output_tokens_per_s": 14539.52687496298,
      "total_tokens_per_s": 29079.05374992596,
      "ttft_p99_ms": 1890.9570557200036,
      "itl_median_ms": 21.60183600005894,
      "accuracy": {
        "mmlu_pro": 0.6318151595744681,
        "gpqa_diamond_cot_zeroshot": 0.494949494949495,
        "gsm8k": 0.8385140257771039,
        "minerva_math500": 0.808,
        "ifeval": 0.833641404805915,
        "arc_challenge_chat": 0.8796928327645052
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.045897871304359926,
          "usd_per_1m_total_tokens": 0.022948935652179963,
          "usd_per_1k_requests": 0.005874927526958071
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.051392471615354735,
          "usd_per_1m_total_tokens": 0.025696235807677367,
          "usd_per_1k_requests": 0.006578236366765406
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.06667647804371302,
          "usd_per_1m_total_tokens": 0.03333823902185651,
          "usd_per_1k_requests": 0.008534589189595267
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.07355428093647425,
          "usd_per_1m_total_tokens": 0.03677714046823712,
          "usd_per_1k_requests": 0.009414947959868704
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.0819604844720713,
          "usd_per_1m_total_tokens": 0.04098024223603565,
          "usd_per_1k_requests": 0.010490942012425127
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.1255199027928924,
          "usd_per_1m_total_tokens": 0.0627599513964462,
          "usd_per_1k_requests": 0.016066547557490228
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.12553900780092786,
          "usd_per_1m_total_tokens": 0.06276950390046393,
          "usd_per_1k_requests": 0.016068992998518765
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 0.20975579372198558,
          "usd_per_1m_total_tokens": 0.10487789686099279,
          "usd_per_1k_requests": 0.026848741596414157
        }
      ]
    },
    {
      "run_id": "20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "bf16",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "prefill-1024-128",
      "input_len": 1024,
      "output_len": 128,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 80,
      "max_users_kv_cache": 327,
      "capacity_is_lower_bound": false,
      "requests_per_s": 23.575623259145114,
      "output_tokens_per_s": 3017.6797771705747,
      "total_tokens_per_s": 27159.117994535172,
      "ttft_p99_ms": 1952.5569503699944,
      "itl_median_ms": 11.956969000038953,
      "accuracy": {
        "mmlu_pro": 0.6318151595744681,
        "gpqa_diamond_cot_zeroshot": 0.494949494949495,
        "gsm8k": 0.8385140257771039,
        "minerva_math500": 0.808,
        "ifeval": 0.833641404805915,
        "arc_challenge_chat": 0.8796928327645052
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.22114120205260346,
          "usd_per_1m_total_tokens": 0.024571244672511493,
          "usd_per_1k_requests": 0.02830607386273324
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.24761481581814151,
          "usd_per_1m_total_tokens": 0.027512757313126836,
          "usd_per_1k_requests": 0.03169469642472211
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.3212549097417524,
          "usd_per_1m_total_tokens": 0.035694989971305824,
          "usd_per_1k_requests": 0.041120628446944314
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.3543929520073773,
          "usd_per_1m_total_tokens": 0.03937699466748637,
          "usd_per_1k_requests": 0.045362297856944296
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.39489500366536323,
          "usd_per_1m_total_tokens": 0.043877222629484805,
          "usd_per_1k_requests": 0.050546560469166495
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.6047692713476542,
          "usd_per_1m_total_tokens": 0.06719658570529491,
          "usd_per_1k_requests": 0.07741046673249974
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.6048613214650587,
          "usd_per_1m_total_tokens": 0.06720681349611764,
          "usd_per_1k_requests": 0.07742224914752752
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 1.010627443995895,
          "usd_per_1m_total_tokens": 0.11229193822176613,
          "usd_per_1k_requests": 0.1293603128314746
        }
      ]
    },
    {
      "run_id": "20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "bf16",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "decode-128-1024",
      "input_len": 128,
      "output_len": 1024,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 396,
      "max_users_kv_cache": 327,
      "capacity_is_lower_bound": false,
      "requests_per_s": 13.009802864979397,
      "output_tokens_per_s": 13322.038133738903,
      "total_tokens_per_s": 14987.292900456265,
      "ttft_p99_ms": 1968.1767526007388,
      "itl_median_ms": 25.77042199936841,
      "accuracy": {
        "mmlu_pro": 0.6318151595744681,
        "gpqa_diamond_cot_zeroshot": 0.494949494949495,
        "gsm8k": 0.8385140257771039,
        "minerva_math500": 0.808,
        "ifeval": 0.833641404805915,
        "arc_challenge_chat": 0.8796928327645052
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.05009243530411984,
          "usd_per_1m_total_tokens": 0.044526609159217634,
          "usd_per_1k_requests": 0.051294653751418715
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.05608918205464634,
          "usd_per_1m_total_tokens": 0.049857050715241186,
          "usd_per_1k_requests": 0.057435322423957845
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.07276997969171589,
          "usd_per_1m_total_tokens": 0.06468442639263634,
          "usd_per_1k_requests": 0.07451645920431707
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.08027633862839718,
          "usd_per_1m_total_tokens": 0.07135674544746416,
          "usd_per_1k_requests": 0.08220297075547871
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.08945077732878541,
          "usd_per_1m_total_tokens": 0.07951180207003149,
          "usd_per_1k_requests": 0.09159759598467626
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.13699105059443362,
          "usd_per_1m_total_tokens": 0.12176982275060767,
          "usd_per_1k_requests": 0.14027883580870004
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.13701190159147994,
          "usd_per_1m_total_tokens": 0.1217883569702044,
          "usd_per_1k_requests": 0.14030018722967547
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 0.2289251816714378,
          "usd_per_1m_total_tokens": 0.20348905037461137,
          "usd_per_1k_requests": 0.2344193860315523
        }
      ]
    },
    {
      "run_id": "20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "bf16",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "balanced-2048-2048",
      "input_len": 2048,
      "output_len": 2048,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 40,
      "max_users_kv_cache": 92,
      "capacity_is_lower_bound": false,
      "requests_per_s": 1.3743562113996448,
      "output_tokens_per_s": 2814.6815209464726,
      "total_tokens_per_s": 5629.363041892945,
      "ttft_p99_ms": 1994.2047643401384,
      "itl_median_ms": 13.249820000055479,
      "accuracy": {
        "mmlu_pro": 0.6318151595744681,
        "gpqa_diamond_cot_zeroshot": 0.494949494949495,
        "gsm8k": 0.8385140257771039,
        "minerva_math500": 0.808,
        "ifeval": 0.833641404805915,
        "arc_challenge_chat": 0.8796928327645052
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 0.23709017463152776,
          "usd_per_1m_total_tokens": 0.11854508731576388,
          "usd_per_1k_requests": 0.48556067764536887
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 0.26547309763520216,
          "usd_per_1m_total_tokens": 0.13273654881760108,
          "usd_per_1k_requests": 0.543688903956894
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 0.344424204738608,
          "usd_per_1m_total_tokens": 0.172212102369304,
          "usd_per_1k_requests": 0.7053807713046693
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 0.37995220293514065,
          "usd_per_1m_total_tokens": 0.18997610146757032,
          "usd_per_1k_requests": 0.778142111611168
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 0.4233753118420138,
          "usd_per_1m_total_tokens": 0.2116876559210069,
          "usd_per_1k_requests": 0.8670726386524443
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 0.6483859670867206,
          "usd_per_1m_total_tokens": 0.3241929835433603,
          "usd_per_1k_requests": 1.3278944605936036
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 0.6484846559705998,
          "usd_per_1m_total_tokens": 0.3242423279852999,
          "usd_per_1k_requests": 1.3280965754277883
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 1.083515124998754,
          "usd_per_1m_total_tokens": 0.541757562499377,
          "usd_per_1k_requests": 2.219038975997448
        }
      ]
    },
    {
      "run_id": "20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
      "experiment": "2026-09-24-qwen3-8b-pipeline",
      "model": "Qwen/Qwen3-8B",
      "precision": "bf16",
      "gpu_type": "H100-80GB",
      "gpu_count": 1,
      "engine_version": "0.30.0",
      "workload": "longctx-8192-512",
      "input_len": 8192,
      "output_len": 512,
      "slo": {
        "ttft_p99_ms": 2000.0,
        "itl_median_ms": 50.0
      },
      "max_users_slo": 7,
      "max_users_kv_cache": 43,
      "capacity_is_lower_bound": false,
      "requests_per_s": 0.9796009355880976,
      "output_tokens_per_s": 501.55567902110596,
      "total_tokens_per_s": 8526.446543358801,
      "ttft_p99_ms": 1573.7907990591339,
      "itl_median_ms": 10.181732499404461,
      "accuracy": {
        "mmlu_pro": 0.6318151595744681,
        "gpqa_diamond_cot_zeroshot": 0.494949494949495,
        "gsm8k": 0.8385140257771039,
        "minerva_math500": 0.808,
        "ifeval": 0.833641404805915,
        "arc_challenge_chat": 0.8796928327645052
      },
      "offers": [
        {
          "provider": "vast",
          "provisioning": "on-demand",
          "product": "H100 SXM, marketplace median",
          "source": "https://console.vast.ai/api/v0/bundles/",
          "checked_at": "2026-10-19",
          "usd_per_hour": 2.4024,
          "usd_per_month": 1753.75,
          "usd_per_1m_output_tokens": 1.3305269210305388,
          "usd_per_1m_total_tokens": 0.07826628947238465,
          "usd_per_1k_requests": 0.6812297835676359
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Community Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 2.69,
          "usd_per_month": 1963.7,
          "usd_per_1m_output_tokens": 1.4898091148735222,
          "usd_per_1m_total_tokens": 0.08763583028667778,
          "usd_per_1k_requests": 0.7627822668152433
        },
        {
          "provider": "runpod",
          "provisioning": "on-demand",
          "product": "H100 SXM, Secure Cloud",
          "source": "https://www.runpod.io/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.49,
          "usd_per_month": 2547.7,
          "usd_per_1m_output_tokens": 1.9328750226425995,
          "usd_per_1m_total_tokens": 0.11369853074368232,
          "usd_per_1k_requests": 0.9896320115930108
        },
        {
          "provider": "nebius",
          "provisioning": "on-demand",
          "product": "HGX H100",
          "source": "https://nebius.com/prices",
          "checked_at": "2026-09-25",
          "usd_per_hour": 3.85,
          "usd_per_month": 2810.5,
          "usd_per_1m_output_tokens": 2.132254681138684,
          "usd_per_1m_total_tokens": 0.12542674594933437,
          "usd_per_1k_requests": 1.0917143967430063
        },
        {
          "provider": "lambda",
          "provisioning": "on-demand",
          "product": "1x H100 SXM",
          "source": "https://lambda.ai/pricing",
          "checked_at": "2026-09-25",
          "usd_per_hour": 4.29,
          "usd_per_month": 3131.7,
          "usd_per_1m_output_tokens": 2.375940930411676,
          "usd_per_1m_total_tokens": 0.13976123120068684,
          "usd_per_1k_requests": 1.2164817563707782
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g",
          "source": "run 20260924T125045Z_h100-80gbx1_qwen3-8b_bf16",
          "checked_at": "2026-09-24",
          "measured": true,
          "usd_per_hour": 6.57,
          "usd_per_month": 4796.1,
          "usd_per_1m_output_tokens": 3.638678767553546,
          "usd_per_1m_total_tokens": 0.2140399275031498,
          "usd_per_1k_requests": 1.8630035289874158
        },
        {
          "provider": "gcp",
          "provisioning": "spot",
          "product": "a3-highgpu-1g Spot, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 6.571,
          "usd_per_month": 4796.83,
          "usd_per_1m_output_tokens": 3.6392325999382575,
          "usd_per_1m_total_tokens": 0.21407250587872104,
          "usd_per_1k_requests": 1.8632870911683879
        },
        {
          "provider": "gcp",
          "provisioning": "on-demand",
          "product": "a3-highgpu-1g, us-central1",
          "source": "https://cloud.google.com/billing/v1/how-tos/catalog-api (Compute Engine SKUs: GPU + vCPU + RAM)",
          "checked_at": "2026-09-26",
          "usd_per_hour": 10.9791,
          "usd_per_month": 8014.74,
          "usd_per_1m_output_tokens": 6.080581134984345,
          "usd_per_1m_total_tokens": 0.3576812432343732,
          "usd_per_1k_requests": 3.1132575411119845
        }
      ]
    }
  ],
  "api": [
    {
      "model": "Qwen/Qwen3-8B",
      "workload": "balanced-2048-2048",
      "provider": "Fireworks",
      "quantization": null,
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "usd_per_1k_requests": 0.8192
    },
    {
      "model": "Qwen/Qwen3-8B",
      "workload": "chat-128-128",
      "provider": "Fireworks",
      "quantization": null,
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "usd_per_1k_requests": 0.0512
    },
    {
      "model": "Qwen/Qwen3-8B",
      "workload": "decode-128-1024",
      "provider": "Fireworks",
      "quantization": null,
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "usd_per_1k_requests": 0.23040000000000002
    },
    {
      "model": "Qwen/Qwen3-8B",
      "workload": "longctx-8192-512",
      "provider": "Fireworks",
      "quantization": null,
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "usd_per_1k_requests": 1.7408000000000003
    },
    {
      "model": "Qwen/Qwen3-8B",
      "workload": "prefill-1024-128",
      "provider": "Fireworks",
      "quantization": null,
      "usd_per_1m_input": 0.2,
      "usd_per_1m_output": 0.2,
      "usd_per_1k_requests": 0.23040000000000002
    }
  ],
  "comparisons": [],
  "openrouter_slugs": {
    "Qwen/Qwen3-8B": "qwen/qwen3-8b",
    "Qwen/Qwen3-14B": "qwen/qwen3-14b",
    "Qwen/Qwen3.8-27B": "qwen/qwen3.8-27b"
  }
}