[whatmodel.sh](../../README.md) / Llama 3.1 Nemotron Ultra 253B v1 # Llama 3.1 Nemotron Ultra 253B v1 NVIDIA · released 2025-04 · 253B params · nvidia-open-model **score 60.7** ±1.4 · #195 of 732 · from 10 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 Llama 3.1 Nemotron Ultra 253B v1 among every model on that task or test, worst to best. 
task            vs the others          rank  counts  tests
agent tasks     ▁▁▃▆●▄▅▅▄▄▂▁   #239 of 313*    4.5%      3
coding          ▁▁▁▃▅▇█▇▅●▃▁    #72 of 469*    1.9%      2
chat & writing  ▁▁▁▃▆▄▆█●▇▇▃    #165 of 402     31%      3
math            ●··········█        #2 of 2     31%      3
agent tasks · #239 of 313 · 4.5% of its score Beats [o3-mini](o3-mini.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Phi-3.5-mini-instruct](phi-3-5-mini-instruct.md), [Jamba 1.5 Mini](jamba-1-5-mini.md) here, though they rank lower overall. 
test                             result  vs the others in its run   difficulty        counts  why
BFCL v3                           56.7%  ▁▁▁▁▁▃▅●▅▄▁▁ beats 64/114  medium · avg 51%    4.5%  blunt · at its level
BFCL V4 (December 2025 archive)   10.0%  ▁▁▁▂▆▆▅▆█▂▄▃ beats 0/114   hard · avg 31%         —  failed run: No function call was ever e…
BFCL v2 Live (maker)              74.1%  █··········● beats 1/1                            —  not counted yet: no other model in the …
coding · #72 of 469 · 1.9% of its score Beats [DeepSeek V4 Flash Vision (Exp)](deepseek-v4-flash-vision-exp.md), [GPT-5.2](gpt-5-2.md), [GPT-5](gpt-5.md) here, though they rank higher overall. 
test                                result  vs the others in its run  difficulty        counts  why
LiveCodeBench official release_v5…   74.9%  █···█····●·█ beats 2/3    medium · avg 51%    1.9%  at its level · noisy
LiveCodeBench 2024-08 to 2025-02 …   66.3%  █··········● beats 1/1                           —  not counted yet: no other model in the …
chat & writing · #165 of 402 · 31% of its score Beats [MiniMax-M2](minimax-m2.md), [MiniCPM5 2B](minicpm5-2b.md), [Granite 4.2 30B](granite-4-2-30b.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)                      1347.6  ▁▁▁▂▃▃▃▆●▇█▃ beats 190/385  avg 1260           22%  blunt
Arena-Hard (version/judge unspeci…   87.0%  ▃▃▂▅▃▆▅█▅▅▂● beats 29/30    easy · avg 70%    6.1%  small run
IFEval (aggregation unspecified) …   88.9%  ▁▂▄▂▂▄▆█●▅▂▁ beats 102/145  easy · avg 80%    3.8%  small run
math · #2 of 2 · 31% of its score Trails [Llama 3.3 Nemotron Super 49B v1](llama-3-3-nemotron-super-49b-v1.md) here, though they rank lower overall. 
test               result  vs the others in its run   difficulty           counts  why
AIME 2025 (maker)   72.5%  ▁·▁▁▁▂▅▆●▇▅▁ beats 70/133  medium · avg 38%        13%  sharp · at its level · small run
AIME 2024 (maker)   80.8%  ▁▁▁▂▁▃▁▅█●█▃ beats 55/85   medium · avg 48%        11%  sharp · small run
MATH-500 (maker)    97.0%  ▁▃▂▃▄▇▅▇●▄▂▁ beats 69/85   very easy · avg 91%    6.8%  maxed out · small run
other tests (2) 
test                                result  vs the others in its run    difficulty      counts  why
GPQA Diamond (maker)                 76.0%  ▁▁▁▁▂▄▅▅●██▄ beats 181/322  hard · avg 46%     17%  sharp · at its level · small run
EQ-Bench 3 (emotional intelligenc…  1139.8  ▁·▂▁▄●▅▃█▄▁▁ beats 17/62    avg 1034           13%  typical
## 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)             6   58% ▇▇▇▇▇▇▇
LMArena leaderboard dataset                     1   22% ▇▇▇
EQ-Bench 3 canonical leaderboard results        1   13% ▇▇
Berkeley Function Calling Leaderboard r…        1    4% ▇
LiveCodeBench official website results          1    2% ▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 60.7
Licences: NVIDIA ([CC BY 4.0](../../licenses/report-2505.00949v5.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · EQ-Bench 3 canonical leaderboard results ([MIT](../../licenses/eqbench3-MIT.txt)) · Berkeley Function Calling Leaderboard result archive ([Apache-2.0](../../licenses/bfcl-result-archive-v4-Apache-2.0.txt)) · LiveCodeBench official website results ([CC-BY-SA-4.0](../../licenses/lcb-sharealike-NOTICE.txt))
Not counted (3) 
source                  test                            score  why
Berkeley Function Cal…  BFCL V4 (December 2025 archiv…  10.0%  failed run: No function call was ever emitted: 0.00% on every call-producing category and 100% only on Irrelevance Detection, which rewards not calling. The 10% overall is just the irrelevance weight.
NVIDIA                  BFCL v2 Live                    74.1%  too few models have it yet
NVIDIA                  LiveCodeBench 2024-08 to 2025…  66.3%  too few models have it yet
## How to run it ### 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 Llama 3.1 Nemotron Ultra 253B v1 (ud-q4_k_xl, needs ~152.3 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF/tree/main)) ```sh llama cli -hf unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF:UD-Q4_K_XL ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF:UD-Q4_K_XL ``` Other versions: ud-q2_k_xl (~76.4 GB) `llama cli -hf unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF:UD-Q2_K_XL` ### 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 Llama 3.1 Nemotron Ultra 253B v1 (ud-q4_k_xl, needs ~152.3 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF/tree/main)) ```sh lms get https://huggingface.co/unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF@UD-Q4_K_XL -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: ud-q2_k_xl (~76.4 GB) `lms get https://huggingface.co/unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF@UD-Q2_K_XL -y` · 5bit (~174.2 GB, Mac only) `lms get https://huggingface.co/LibraxisAI/Llama-3_1-Nemotron-Ultra-253B-v1-MLX-Q5 -y` ### 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 Llama 3.1 Nemotron Ultra 253B v1 (5bit, needs ~174.2 GB, Mac only, from [huggingface.co/LibraxisAI](https://huggingface.co/LibraxisAI/Llama-3_1-Nemotron-Ultra-253B-v1-MLX-Q5/tree/3e772160e672b5a731c14b8be56a5102557daa5d)) ```sh mlx_lm.chat --model LibraxisAI/Llama-3_1-Nemotron-Ultra-253B-v1-MLX-Q5 ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model LibraxisAI/Llama-3_1-Nemotron-Ultra-253B-v1-MLX-Q5 ``` Model repository and documentation: [huggingface.co/nvidia](https://huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1) [interactive version](https://whatmodel.sh/model/llama-3-1-nemotron-ultra-253b) · [JSON](https://whatmodel.sh/api?format=json&all=1)