[whatmodel.sh](../../README.md) / Llama 3.3 Nemotron Super 49B v1 # Llama 3.3 Nemotron Super 49B v1 NVIDIA · released 2025-03 · 49.9B params · nvidia-open-model-license **score 58.1** ±1.6 · #224 of 732 · from 7 benchmark results across 2 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks Llama 3.3 Nemotron Super 49B v1 among every model on that task or test, worst to best. 
task            vs the others            rank  counts  tests
agent tasks                    not ranked yet       —      1
coding                         not ranked yet       —      1
chat & writing  ▁▁▁▃▆▄▆●▆▇▇▃      #188 of 402     39%      3
math            █··········●          #1 of 2     40%      3
agent tasks · 0.0% of its score 
test                  result  vs the others in its run  counts  why
BFCL v2 Live (maker)   73.7%  ●··········█ beats 0/1         —  not counted yet: no other model in the …
coding · 0.0% of its score 
test                                result  vs the others in its run  counts  why
LiveCodeBench 2024-08 to 2025-02 …   45.5%  ●··········█ beats 0/1         —  not counted yet: no other model in the …
chat & writing · #188 of 402 · 39% of its score Beats [Seed-OSS 36B Instruct](seed-oss-36b.md), [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Nemotron-Cascade-14B-Thinking](nemotron-cascade-14b-thinking.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md), [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md), [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md) here, though they rank lower overall. 
test                                result  vs the others in its run    difficulty      counts  why
LMArena (text)                      1327.7  ▁▁▁▂▃▃▃●▆▇█▃ beats 166/385  avg 1260           26%  blunt
Arena-Hard (version/judge unspeci…   88.3%  ▃▃▂▅▃▆▅█▅▅▂● beats 30/30    easy · avg 70%    7.8%  small run
IFEval (aggregation unspecified) …   89.2%  ▁▂▄▂▂▄▆█●▅▂▁ beats 108/145  easy · avg 80%    4.9%  small run
math · #1 of 2 · 40% of its score Beats [Llama 3.1 Nemotron Ultra 253B v1](llama-3-1-nemotron-ultra-253b.md) here, though they rank higher overall. 
test               result  vs the others in its run   difficulty           counts  why
AIME 2025 (maker)   60.0%  ▁·▁▁▁▂▅●█▇▅▁ beats 47/133  medium · avg 38%        17%  sharp · at its level · small run
AIME 2024 (maker)   67.5%  ▁▁▁▂▁▃▁▅●██▃ beats 32/85   medium · avg 48%        14%  sharp · at its level · small run
MATH-500 (maker)    96.6%  ▁▃▂▃▄▇▅▇●▄▂▁ beats 63/85   very easy · avg 91%    8.9%  maxed out · small run
other tests (1) 
test                  result  vs the others in its run    difficulty      counts  why
GPQA Diamond (maker)   66.7%  ▁▁▁▁▂▄▅▅●██▄ beats 143/322  hard · avg 46%     21%  sharp · at its level · 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 (submitted / maker-reported)        6   73% ▇▇▇▇▇▇▇▇▇
LMArena leaderboard dataset                1   26% ▇▇▇
prior: pull toward the middle                   1% ▇
weighted average                              = 58.1
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/))
Not counted (2) 
source  test                            score  why
NVIDIA  BFCL v2 Live                    73.7%  too few models have it yet
NVIDIA  LiveCodeBench 2024-08 to 2025…  45.5%  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 3.3 Nemotron Super 49B v1 (Q4_K_M, needs ~30.22 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF/tree/6c679d12e88815cf6c66cf627b46ea66acfdeb4d)) ```sh llama cli -hf unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF:Q4_K_M ``` Other versions: ud-q2_k_xl (~19.31 GB) `llama cli -hf unsloth/Llama-3_3-Nemotron-Super-49B-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 3.3 Nemotron Super 49B v1 (Q4_K_M, needs ~30.22 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF/tree/6c679d12e88815cf6c66cf627b46ea66acfdeb4d)) ```sh lms get https://huggingface.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF@Q4_K_M -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: ud-q2_k_xl (~19.31 GB) `lms get https://huggingface.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF@UD-Q2_K_XL -y` · 4bit (~28.1 GB, Mac only) `lms get https://huggingface.co/mlx-community/Llama-3_3-Nemotron-Super-49B-v1-mlx-4bit -y` · 8bit (~56.1 GB, Mac only) `lms get https://huggingface.co/sjug/Llama-3_3-Nemotron-Super-49B-v1-8bit -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 Llama 3.3 Nemotron Super 49B v1 (Q4_K_M, needs ~30.22 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF/tree/6c679d12e88815cf6c66cf627b46ea66acfdeb4d)) ```sh ollama run hf.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF:Q4_K_M ``` Other versions: ud-q2_k_xl (~19.31 GB) `ollama run hf.co/unsloth/Llama-3_3-Nemotron-Super-49B-v1-GGUF:UD-Q2_K_XL` ### 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 3.3 Nemotron Super 49B v1 (4bit, needs ~28.1 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/Llama-3_3-Nemotron-Super-49B-v1-mlx-4bit/tree/e460111cccd41055a1a59291bd05f3b676ed7ca0)) ```sh mlx_lm.chat --model mlx-community/Llama-3_3-Nemotron-Super-49B-v1-mlx-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model mlx-community/Llama-3_3-Nemotron-Super-49B-v1-mlx-4bit ``` Other versions: 8bit (~56.1 GB, Mac only) `mlx_lm.chat --model sjug/Llama-3_3-Nemotron-Super-49B-v1-8bit` Model repository and documentation: [huggingface.co/nvidia](https://huggingface.co/nvidia/Llama-3_3-Nemotron-Super-49B-v1) [interactive version](https://whatmodel.sh/model/llama-3-3-nemotron-super-49b-v1) · [JSON](https://whatmodel.sh/api?format=json&all=1)