[whatmodel.sh](../../README.md) / Nemotron 3 Nano 30B-A3B # Nemotron 3 Nano 30B-A3B NVIDIA · released 2025-12 · 31.6B params (3.6B active) · nvidia-open-model **score 57.0** ±0.9 · #245 of 732 · from 22 benchmark results across 9 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 Nano 30B-A3B among every model on that task or test, worst to best. 
task            vs the others            rank  counts  tests
agent tasks     ▁▁▃▆●▄▅▅▄▄▂▁     #187 of 313*     16%      5
coding          ▁▁▁▃▅▇●▇▅▄▃▁      #262 of 469    9.1%      5
chat & writing  ▁▁▁▃▆▄▆●▆▇▇▃     #215 of 402*    7.8%      2
math            ▁▁▃▅█▆▄●▄▂▁▁      #150 of 431     23%      4
web search                     not ranked yet       —      1
agent tasks · #187 of 313 · 16% of its score Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o3-mini](o3-mini.md), [DeepSeek R1 0528](deepseek-r1-0528.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Jamba 1.5 Large](jamba-1-5-large.md) here, though they rank lower overall. 
test                                result  vs the others in its run   difficulty            counts  why
τ²-bench Telecom (maker)             42.2%  ▁·▅█●▅▂▃▃▂▃▁ beats 35/89   hard · avg 19%           12%  sharp · at its level
Terminal-Bench Hard (maker)           8.5%  ▂▅▆▂▂●▂▂█▅▅▂ beats 11/27   very hard · avg 5.2%    4.3%  sharp · small run
BFCL V4 (December 2025 archive) (…   53.8%  ▁▁▁▂▆▆▅▆●▂▄▃ beats 81/114  hard · avg 31%             —  not counted yet: no other model in the …
Humanity's Last Exam with tools (…   15.5%  ▁·▁●▂▁▄▄▇█▄▅ beats 2/62    hard · avg 15%             —  not counted yet: no other model in the …
τ²-bench aggregate (maker)           49.0%  ▁▁▁·▁●▃▃▃▄█▁ beats 6/39    medium · avg 45%           —  not counted yet: no other model in the …
coding · #262 of 469 · 9.1% of its score Beats [Qwen3-VL 32B Instruct](qwen3-vl-32b-instruct.md), [Solar Open 100B (102B-A12B)](solar-open-100b.md), [GLM-4.7-Flash](glm-4-7-flash.md) here, though they rank higher overall. Trails [DeepSeek LLM 67B Chat](deepseek-llm-67b-chat.md), [QwQ-32B-Preview](qwq-32b-preview.md), [InternLM2.5 20B Chat](internlm2-5-20b-chat.md) here, though they rank lower overall. 
test                                result  vs the others in its run  difficulty            counts  why
LiveCodeBench Pro 2025Q2 (Easy) (…   54.5%  ▂··▃▂▃●▅█·▃▂ beats 8/20   medium · avg 47%        3.8%  at its level
LiveCodeBench Pro 2025Q2 (Medium)…    3.5%  █··▄▂●██▄·▄▆ beats 7/22   very hard · avg 3.5%    3.2%  sharp
SciCode (level unspecified) (make…   33.3%  ▁····▃▃▄●█▆▂ beats 18/40  hard · avg 19%          2.0%  at its level · small run
LiveCodeBench v6 (Aug 2024–May 20…   68.3%  ▂▅▅●█▃▂····▂ beats 8/17   medium · avg 45%           —  not counted yet: no other model in the …
SciCode subproblems (maker)          33.3%  ●··▃▃▃·▃▃█·▅ beats 0/10   hard · avg 19%             —  not counted yet: no other model in the …
chat & writing · #215 of 402 · 7.8% of its score Beats [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Granite 4.2 8B](granite-4-2-8b.md), [Mistral Small 4 (119B-A6.5B)](mistral-small-4.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md), [Granite 4.0 H Micro 3B](granite-4-0-h-micro.md), [Gemma 3n E4B](gemma-3n-e4b.md) here, though they rank lower overall. 
test                             result  vs the others in its run    difficulty      counts  why
LMArena (text)                   1313.6  ▁▁▁▂▃▃▃●▆▇█▃ beats 143/385  avg 1260          7.8%  blunt
Arena-Hard v2 aggregate (maker)   67.7%  ▁▁··▄▃▁▃●▁▁▇ beats 14/22    hard · avg 28%       —  not counted yet: no other model in the …
math · #150 of 431 · 23% of its score Beats [Claude Sonnet 4.6](claude-sonnet-4-6.md), [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md), [MiniMax-M2.7](minimax-m2-7.md) here, though they rank higher overall. 
test                        result  vs the others in its run    difficulty        counts  why
AIME 2025 (maker)            89.1%  ▁·▁▁▁▂▅▆█●▅▁ beats 111/133  medium · avg 64%    8.1%  sharp
AIME 2026 (maker)            89.9%  ▁▁▁▂▂▂●█▆▂·▁ beats 23/52    easy · avg 66%      6.5%  sharp
IMO-AnswerBench (maker)      70.4%  ▁▁·▁▁▁●▆█▄▂▃ beats 11/44    medium · avg 43%    5.6%  at its level
HMMT February 2025 (maker)   84.6%  ▁▁▁▃▁▅▅●██▁▄ beats 15/31    easy · avg 72%      3.2%  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 (…   15.5%  ▁·▁●▂▁▄▄▇█▄▅ beats 2/62   hard · avg 15%       —  not counted yet: no other model in the …
other tests (15) 
test                                result  vs the others in its run    difficulty            counts  why
Medmarks-V mean win rate (19 veri…  0.5114  ▁▄▃▃▇▄●▃▅▁▁▁ beats 19/30    avg 0.418                14%  sharp
Medmarks-V weighted mean win rate…   51.1%  ▂▃▅▂▆▃█●█▅▃▆ beats 19/34    medium · avg 41%        6.4%  blunt · at its level
CPTUB average (SpeakLeash, 0-5 ju…   64.9%  ▁▂▂▂▃█●▆▅▄▂▃ beats 43/95    medium · avg 65%        4.8%  typical
EuroEval Life in the UK (English …   93.0%  ▂·▂▂·▂·▃▃·●▃ beats 9/14     easy · avg 80%          4.1%  typical
MMLU-ProX (maker)                    59.5%  ▁·▁▁●▂▁▃▃█▅▄ beats 3/36     medium · avg 54%        3.8%  at its level · small run · several runs
Polish EQ-Bench v2 (SpeakLeash)      45.8%  ▁··▁▁▂▂●▅▆█▆ beats 17/70    medium · avg 62%        3.6%  typical
EuroEval HellaSwag (English), val…   69.2%  ▂··▂▂·●▆▆█·▄ beats 3/15     medium · avg 63%        2.9%  at its level
MultiChallenge (maker)               38.5%  ▁·▁·▁●▆▅▃▂█▅ beats 3/30     hard · avg 28%          2.0%  at its level · small run
CL-bench (maker)                     12.0%  ●··········▄ beats 0/2      very hard · avg 7.2%    1.1%  blunt · small run
miniF2F (pass@1) (maker)             50.0%  ▃·█▅·····●·▃ beats 7/8      very hard · avg 14%     0.2%  blunt · noisy · small run
Vectara hallucination leaderboard…   90.4%  ▁▂▁▂▅●▂▃▄▃▁▁ beats 48/98    very easy · avg 90%     0.0%  blunt
MMLU-Pro (maker)                     78.3%  ▁▁·▁▁▁▃▂▄●█▁ beats 116/188  medium · avg 54%           —  not counted yet: no other model in the …
Humanity's Last Exam (maker)         10.6%  ▁▁▁▄●▆▇██▆▁▃ beats 22/111   very hard · avg 3.8%       —  not counted yet: no other model in the …
GPQA (subset unspecified) (maker)    73.0%  ▂▁▂▂▄▃▆▃▅●▃▃ beats 83/106   very hard · avg 32%        —  not counted yet: no other model in the …
IFBench prompt-level (maker)         71.5%  ●▄▄··▄·▄···▄ beats 0/6                                 —  not counted yet: no other model in the …
## 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 (submitted / maker-reported)            10   38% ▇▇▇▇▇
Medmarks: A Comprehensive Open-Source L…        2   21% ▇▇
NVIDIA vendor-reported benchmark scores…        2   16% ▇▇
LMArena leaderboard dataset                     1    8% ▇
EuroEval leaderboards                           2    7% ▇
SpeakLeash CPTUB                                1    5% ▇
SpeakLeash Polish EQ-Bench leaderboard          1    4% ▇
NVIDIA (submitted / maker-reported)             2    2% ▇
Vectara hallucination leaderboard               1    0% ▇
prior: pull toward the middle                        0% ▇
weighted average                                   = 57.0
Licences: NVIDIA ([CC BY 4.0](../../licenses/report-2603.19220v2.txt)) · Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks ([CC BY 4.0](../../licenses/paper-2605.01417v1-CC-BY-4.0.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · SpeakLeash CPTUB ([MIT](../../licenses/hf-leaderboard-spaces-MIT.txt)) · SpeakLeash Polish EQ-Bench leaderboard ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt)) · NVIDIA ([CC BY 4.0](../../licenses/report-2512.20848v1.txt)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
Not counted (12) 
source                  test                            score  why
nvidia/NVIDIA-Nemotro…  MMLU-Pro                        78.3%  too few models have it yet
nvidia/NVIDIA-Nemotro…  AIME 2025                       89.1%  only model in this source's run
nvidia/NVIDIA-Nemotro…  GPQA (subset unspecified)       73.0%  too few models have it yet
nvidia/NVIDIA-Nemotro…  SciCode subproblems             33.3%  too few models have it yet
nvidia/NVIDIA-Nemotro…  Humanity's Last Exam            10.6%  too few models have it yet
nvidia/NVIDIA-Nemotro…  τ²-bench aggregate              49.0%  too few models have it yet
nvidia/NVIDIA-Nemotro…  BFCL V4 (December 2025 archiv…  53.8%  too few models have it yet
nvidia/NVIDIA-Nemotro…  IFBench prompt-level            71.5%  too few models have it yet
nvidia/NVIDIA-Nemotro…  Arena-Hard v2 aggregate         67.7%  too few models have it yet
NVIDIA                  Humanity's Last Exam with too…  15.5%  too few models have it yet
NVIDIA                  LiveCodeBench v6 (Aug 2024–Ma…  68.3%  too few models have it yet
NVIDIA                  MultiChallenge                  38.5%  only model in this source's run
## How to run it ### Through an API api (Deep Infra) · $0.088/M blended · $0.05 in / $0.2 out per 1M tokens · [deepinfra.com](https://deepinfra.com/models) [price source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/providers/deepinfra/models/nvidia/Nemotron-3-Nano-30B-A3B.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-nano-30b-a3b.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 Nano 30B-A3B (Q4_K_M, needs ~22.4 GB, from [huggingface.co/ggml-org](https://huggingface.co/ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF/tree/f9d9b441a049bf27473c3cdd9cec220f5657b862)) ```sh llama cli -hf ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF:Q4_K_M ``` Other versions: Q8_0 (~33.6 GB) `llama cli -hf ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-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 Nano 30B-A3B (Q4_K_M, needs ~22.4 GB, from [huggingface.co/ggml-org](https://huggingface.co/ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF/tree/f9d9b441a049bf27473c3cdd9cec220f5657b862)) ```sh lms get https://huggingface.co/ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-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 (~33.6 GB) `lms get https://huggingface.co/ggml-org/NVIDIA-Nemotron-3-Nano-30B-A3B-GGUF@Q8_0 -y` · 4bit (~17.8 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-4bit -y` · 8bit (~33.6 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-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 Nano 30B-A3B (Q4_K_M, needs ~19.46 GB, from [ollama.com](https://ollama.com/library/nemotron-3-nano/tags)) ```sh ollama run nemotron-3-nano:30b ``` Other versions: Q8_0 (~38.42 GB) `ollama run nemotron-3-nano:30b-a3b-q8_0` ### 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 Nano 30B-A3B (4bit, needs ~17.8 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-4bit/tree/543873e66e36362182738b0d2863a766728045aa)) ```sh mlx_lm.chat --model lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-4bit ``` Other versions: 8bit (~33.6 GB, Mac only) `mlx_lm.chat --model lmstudio-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-8bit` Model repository and documentation: [huggingface.co/nvidia](https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-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
AMD Ryzen AI Max+ 395 (Stri…  IQ4_XS      75.97 tok/s  llama.cpp 2016bf2 (build_number 1…
Apple M5 Max (40-core GPU),…  q4_k_m     104.78 tok/s  ollama (version not recorded) · l…
Each speed links to its source, which has the full settings. ## Quantization evidence How much of the full-precision result a quantized version kept, measured within one study. Rows from the same study share a table and aren't independent. They don't describe every download of that precision. 
precision                       test                                full → quantized   kept
NVFP4 PTQ (post-training, W4A…  GPQA-Diamond accuracy (NVIDIA eva…      0.73 → 0.716  98.1%
NVFP4 PTQ (post-training, W4A…  LiveCodeBench v5 accuracy (NVIDIA…     0.721 → 0.689  95.6%
NVFP4 PTQ (post-training, W4A…  AIME 2025 accuracy (NVIDIA evalua…      0.891 → 0.85  95.4%
NVFP4 QAD (quantization-aware…  GPQA-Diamond accuracy (NVIDIA eva…      0.73 → 0.727  99.6%
NVFP4 QAD (quantization-aware…  AIME 2025 accuracy (NVIDIA evalua…     0.891 → 0.879  98.7%
NVFP4 QAD (quantization-aware…  LiveCodeBench v5 accuracy (NVIDIA…     0.721 → 0.689  95.6%
Each kept value links to its source. [interactive version](https://whatmodel.sh/model/nemotron-3-nano-30b-a3b) · [JSON](https://whatmodel.sh/api?format=json&all=1)