[whatmodel.sh](../../README.md) / Nemotron 3 Super 120B-A12B # Nemotron 3 Super 120B-A12B NVIDIA · released 2026-03 · 120.6B params (12.7B active) · nvidia-open-model **score 63.6** ±0.8 · #160 of 732 · from 26 benchmark results across 5 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks Nemotron 3 Super 120B-A12B among every model on that task or test, worst to best. 
task            vs the others            rank  counts  tests
agent tasks     ▁▁▃▆█▄●▅▄▄▂▁      #141 of 313     25%     10
coding          ▁▁▁▃▅▇█▇●▄▃▁      #125 of 469     17%      7
chat & writing  ▁▁▁▃▆▄▆█●▇▇▃     #154 of 402*    6.8%      1
math            ▁▁▃▅█▆▄●▄▂▁▁      #133 of 431     36%      7
web search                     not ranked yet       —      2
agent tasks · #141 of 313 · 25% of its score Beats [o3](o3.md), [Gemini 2.5 Pro](gemini-2-5-pro.md), [Gemini 3.1 Flash-Lite](gemini-3-1-flash-lite.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Large](jamba-1-5-large.md), [GPT-4o (2024-05-13)](gpt-4o-2024-05-13.md) here, though they rank lower overall. 
test                                result  vs the others in its run   difficulty            counts  why
τ²-bench Telecom (maker)             64.4%  ▁·▅█▆●▂▃▃▂▃▁ beats 51/89   hard · avg 19%           10%  sharp · at its level
SWE-bench Verified (maker)           59.5%  ▁▁▁▁▁▁▂▄●▆▁▁ beats 55/138  hard · avg 22%          5.5%  sharp · at its level · small run
Terminal-Bench Hard (maker)          25.8%  ▂▅▆▂▂▆▂▂●▅▅▂ beats 20/27   very hard · avg 5.2%    5.3%  sharp · small run
τ²-bench Retail (maker)              64.3%  ▁····▂▁▃●█▃▁ beats 16/44   medium · avg 35%        2.3%  at its level · small run
τ²-bench Airline (maker)             56.9%  ▁▁·▂▁▃▂▁●▅▂▃ beats 20/42   hard · avg 29%          1.8%  at its level · small run
SWE-bench Multilingual (maker)       45.8%  ▁··▁▃●▅██▂▂▁ beats 11/65   hard · avg 18%             —  not counted yet: no other model in the …
Humanity's Last Exam with tools (…   22.8%  ▁·▁▂●▁▄▄▇█▄▅ beats 9/62    hard · avg 15%             —  not counted yet: no other model in the …
BrowseComp (maker)                   31.3%  ▂·▃▁▂●▆▅█▂▂▄ beats 11/46   very hard · avg 11%        —  not counted yet: no other model in the …
τ²-bench aggregate (maker)           61.2%  ▁▁▁·▁▁●▃▃▄█▁ beats 11/39   medium · avg 45%           —  not counted yet: no other model in the …
Terminal-Bench 2.0 (maker)           31.0%  ▂▂▅●████▆▂▃▃ beats 6/36    very hard · avg 10%        —  not counted yet: no other model in the …
coding · #125 of 469 · 17% of its score Beats [Qwen3.5 Plus](qwen3-5-plus.md), [Qwen3-Coder-Next (80B-A3B)](qwen3-coder-next.md), [Gemini 2.5 Pro](gemini-2-5-pro.md) here, though they rank higher overall. Trails [Llama 3.1 Nemotron Ultra 253B v1](llama-3-1-nemotron-ultra-253b.md), [Solar Pro 4](solar-pro-4.md) here, though they rank lower overall. 
test                                result  vs the others in its run   difficulty            counts  why
SWE-bench Verified (maker)           59.5%  ▁▁▁▁▁▁▂▄●▆▁▁ beats 55/138  hard · avg 22%          5.5%  sharp · at its level · small run
LiveCodeBench Pro 2025Q2 (Medium)…   23.2%  █··▄▂▂██▄·▄● beats 20/22   very hard · avg 3.5%    4.8%  sharp
LiveCodeBench Pro 2025Q2 (Easy) (…   81.7%  ▂··▃▂▃▆▅█·●▂ beats 19/20   medium · avg 47%        2.7%  typical
SciCode (level unspecified) (make…   42.1%  ▁····▃▃▄▅█●▂ beats 32/40   hard · avg 19%          2.0%  at its level · small run
SciCode subproblems (maker)          42.3%  ▃··▃▃●·▃▃█·▅ beats 3/10    hard · avg 21%          1.8%  at its level · small run
SWE-bench Multilingual (maker)       45.8%  ▁··▁▃●▅██▂▂▁ beats 11/65   hard · avg 18%             —  not counted yet: no other model in the …
LiveCodeBench v5 (Jul–Dec 2024) (…   82.1%  █··········● beats 1/1                                —  not counted yet: no other model in the …
chat & writing · #154 of 402 · 6.8% of its score Beats [INTELLECT-3](intellect-3.md), [MiniMax-M2](minimax-m2.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md), [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md), [LFM2.5 2.6B](lfm2-5-2-6b.md) here, though they rank lower overall. 
test            result  vs the others in its run    difficulty  counts  why
LMArena (text)  1359.9  ▁▁▁▂▃▃▃▆●▇█▃ beats 203/385  avg 1260      6.8%  blunt
math · #133 of 431 · 36% of its score Beats [Claude Sonnet 4.6](claude-sonnet-4-6.md), [MiniMax-M3](minimax-m3.md), [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md) here, though they rank higher overall. Trails [Olmo 3.1 32B Think](olmo-3-1-32b-think.md), [Qwen3-VL 30B-A3B Thinking](qwen3-vl-30b-a3b-thinking.md), [Granite 4.2 8B](granite-4-2-8b.md) here, though they rank lower overall. 
test                                result  vs the others in its run    difficulty            counts  why
AIME 2025 (maker)                    90.2%  ▁·▁▁▁▂▅▆█▇●▁ beats 121/133  medium · avg 64%        9.3%  sharp · several runs
MathArena ArXivMath February 2026…   30.5%  ▁▁▁▁●▇▄▃▁▃·▃ beats 5/24     very hard · avg 9.4%    6.7%  at its level
MathArena ArXivMath January 2026 …   48.9%  ▂··▄▂●█▆▆▄▄█ beats 4/23     hard · avg 23%          5.9%  at its level
IMO-AnswerBench (maker)              77.2%  ▁▁·▁▁▁▄●█▄▂▃ beats 19/44    medium · avg 43%        4.1%  typical
AIME 2026 (maker)                    89.8%  ▁▁▁▂▂▂●█▆▂·▁ beats 22/52    easy · avg 66%          3.8%  sharp
MathArena ArXivMath December 2025…   33.8%  ▃·●▃█▁·▄·▁▁▁ beats 3/18     hard · avg 19%          3.7%  at its level
HMMT February 2025 (maker)           94.2%  ▁▁▁▃▁▅▅▃█●▁▄ beats 24/31    easy · avg 75%          2.7%  maxed out · small run
web search · 0.0% of its score 
test                                result  vs the others in its run  difficulty           counts  why
Humanity's Last Exam with tools (…   22.8%  ▁·▁▂●▁▄▄▇█▄▅ beats 9/62   hard · avg 15%            —  not counted yet: no other model in the …
BrowseComp (maker)                   31.3%  ▂·▃▁▂●▆▅█▂▂▄ beats 11/46  very hard · avg 11%       —  not counted yet: no other model in the …
other tests (8) 
test                               result  vs the others in its run    difficulty            counts  why
GPQA (subset unspecified) (maker)   80.5%  ▂▁▂▂▄▃▆▃▅█●▃ beats 96/106   hard · avg 48%          6.2%  sharp · at its level
Humanity's Last Exam (maker)        18.5%  ▁▁▁▄▆▆●██▆▁▃ beats 51/111   very hard · avg 5.5%    3.4%  small run
MMLU-Pro (maker)                    83.8%  ▁▁·▁▁▁▃▂▄▅●▁ beats 161/188  medium · avg 59%        2.5%  small run
WideSearch corrected dataset        19.5%  ●······▃▃▃█▃ beats 0/7      hard · avg 22%          2.1%  at its level · noisy
MultiChallenge (maker)              55.2%  ▁·▁·▁▃▆▅▃●█▅ beats 18/30    hard · avg 28%          1.8%  at its level · small run
SkillsBench 87-task subset          20.3%  ▄··●▄▄···█·█ beats 1/7      very hard · avg 9.1%    1.7%  at its level
MMLU-ProX (maker)                   79.4%  ▁·▁▁▁▂▁▃▃●▅▄ beats 22/36    medium · avg 54%        1.5%  at its level · small run
IFBench prompt-level (maker)        73.4%  █●▄··▄·▄···▄ beats 2/6      medium · avg 56%        1.1%  small run
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
NVIDIA vendor-reported benchmark scores…       12   47% ▇▇▇▇▇▇
NVIDIA (submitted / maker-reported)             8   26% ▇▇▇
MathArena ArXivMath / BrokenArXiv / ArX…        3   16% ▇▇
LMArena leaderboard dataset                     1    7% ▇
inclusionAI/Ling-3.0-flash model card           2    4% ▇
prior: pull toward the middle                        0% ▇
weighted average                                   = 63.6
Licences: NVIDIA ([CC BY 4.0](../../licenses/report-2603.19220v2.txt)) · MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets ([CC BY-SA 4.0](../../licenses/matharena-sharealike-NOTICE.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · inclusionAI/Ling-3.0-flash model card ([MIT](../../licenses/cards-final-NOTICE.txt))
Not counted (16) 
source                  test                            score  why
nvidia/NVIDIA-Nemotro…  MMLU-Pro                        83.7%  only model in this source's run
nvidia/NVIDIA-Nemotro…  AIME 2025                       90.2%  only model in this source's run
nvidia/NVIDIA-Nemotro…  HMMT February 2025              93.7%  only model in this source's run
nvidia/NVIDIA-Nemotro…  GPQA (subset unspecified)       79.2%  only model in this source's run
nvidia/NVIDIA-Nemotro…  LiveCodeBench v5 (Jul–Dec 202…  81.2%  too few models have it yet
nvidia/NVIDIA-Nemotro…  SciCode subproblems             42.0%  only model in this source's run
nvidia/NVIDIA-Nemotro…  Humanity's Last Exam            18.3%  only model in this source's run
nvidia/NVIDIA-Nemotro…  Terminal-Bench 2.0              31.0%  too few models have it yet
nvidia/NVIDIA-Nemotro…  SWE-bench Multilingual          45.8%  too few models have it yet
nvidia/NVIDIA-Nemotro…  τ²-bench aggregate              61.2%  too few models have it yet
nvidia/NVIDIA-Nemotro…  IFBench prompt-level            72.6%  only model in this source's run
NVIDIA                  BrowseComp                      31.3%  too few models have it yet
NVIDIA                  Humanity's Last Exam with too…  22.8%  too few models have it yet
NVIDIA                  MMLU-ProX                       79.4%  only model in this source's run
NVIDIA                  MultiChallenge                  55.2%  only model in this source's run
NVIDIA vendor-reporte…  LiveCodeBench v5 (Jul–Dec 202…  82.1%  too few models have it yet
## How to run it ### Through an API api (Amazon Bedrock) · $0.28/M blended · $0.15 in / $0.65 out per 1M tokens · [docs.aws.amazon.com](https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html) [price source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/providers/amazon-bedrock/models/nvidia.nemotron-super-3-120b.toml): Standard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ. · [context source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/models/nvidia/nemotron-3-super-120b-a12b.toml) ### llama.cpp (recommended) 1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md) ```sh # macOS / Linux curl -LsSf https://llama.app/install.sh | sh # Windows irm https://llama.app/install.ps1 | iex ``` 2. Run Nemotron 3 Super 120B-A12B (Q4_K_M, needs ~86.1 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF/tree/16410500b9bdc62081c34cc2dff52471e001f970)) ```sh llama cli -hf lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:Q4_K_M ``` Other versions: Q8_0 (~128.5 GB) `llama cli -hf lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF:Q8_0` ### LM Studio 1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download) 2. Run Nemotron 3 Super 120B-A12B (Q4_K_M, needs ~86.1 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF/tree/16410500b9bdc62081c34cc2dff52471e001f970)) ```sh lms get https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@Q4_K_M -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: Q8_0 (~128.5 GB) `lms get https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Super-120B-A12B-GGUF@Q8_0 -y` · 4bit (~68 GB, Mac only) `lms get https://huggingface.co/mlx-community/NVIDIA-Nemotron-3-Super-120B-A12B-4bit -y` · 8bit (~128.3 GB, Mac only) `lms get https://huggingface.co/mlx-community/NVIDIA-Nemotron-3-Super-120B-A12B-8bit -y` ### Ollama 1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download) ```sh # Linux curl -fsSL https://ollama.com/install.sh | sh ``` 2. Run Nemotron 3 Super 120B-A12B (Q4_K_M, needs ~72.86 GB, from [ollama.com](https://ollama.com/library/nemotron-3-super/tags)) ```sh ollama run nemotron-3-super:120b ``` ### MLX (Mac) 1. Install MLX (Mac). Apple silicon Macs only, text only. ```sh # with uv uv tool install mlx-lm # or pip install mlx-lm ``` 2. Run Nemotron 3 Super 120B-A12B (4bit, needs ~68 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/NVIDIA-Nemotron-3-Super-120B-A12B-4bit/tree/46d2ec99579734e0da13b6189bb7ea500b7a1e00)) ```sh mlx_lm.chat --model mlx-community/NVIDIA-Nemotron-3-Super-120B-A12B-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model mlx-community/NVIDIA-Nemotron-3-Super-120B-A12B-4bit ``` Other versions: 8bit (~128.3 GB, Mac only) `mlx_lm.chat --model mlx-community/NVIDIA-Nemotron-3-Super-120B-A12B-8bit` Model repository and documentation: [huggingface.co/nvidia](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16) ## Published speed measurements Reference results for the hardware, engine and settings below. These are not verified speeds of the downloads above or predictions for your machine. 
hardware                      precision               speed  engine and settings
NVIDIA DGX Spark (GB10)       Q4_K (ggml-org …  19.94 tok/s  llama.cpp 04a65daab (8268), CUDA …
NVIDIA DGX Spark (GB10)       Q4_K (ggml-org …  19.95 tok/s  llama.cpp 04a65daab (8268), CUDA …
NVIDIA DGX Spark (GB10)       Q4_K (ggml-org …  19.83 tok/s  llama.cpp 04a65daab (8268), CUDA …
NVIDIA DGX Spark (GB10)       Q4_K (ggml-org …  19.84 tok/s  llama.cpp 04a65daab (8268), CUDA …
NVIDIA DGX Spark (GB10)       Q4_K (ggml-org …  19.45 tok/s  llama.cpp 04a65daab (8268), CUDA …
AMD Ryzen AI Max+ 395 (Stri…  UD-IQ4_XS         18.24 tok/s  llama.cpp ac4cddeb0 (build_number…
AMD Ryzen AI Max+ 395 (Stri…  UD-IQ4_XS         18.43 tok/s  llama.cpp 2016bf2 (build_number 1…
AMD Ryzen AI Max+ 395 (Stri…  UD-IQ4_XS         18.93 tok/s  llama.cpp 98d5e8ba8 (build_number…
Each speed links to its source, which has the full settings. [interactive version](https://whatmodel.sh/model/nemotron-3-super-120b-a12b) · [JSON](https://whatmodel.sh/api?format=json&all=1)