whatmodel.sh / Llama 3.1 Nemotron Ultra 253B v1
Llama 3.1 Nemotron Ultra 253B v1
↗ share
+ contribute
Llama 3.1 Nemotron Ultra 253B v1 is an open-weights language model from NVIDIA, released April 2025. Its published size is 253B parameters and a 128k-token context window. Licence: nvidia-open-model.
$ how it compares
The LLMs ranked either side of Llama 3.1 Nemotron Ultra 253B v1, out of 828 with a score, then more from NVIDIA.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 218 |  Grok 4 Fast | 61±2 | comparison only | ||
| 219 |  GLM-4.5V | 61±2 | api $0.90/M +local | ||
| 220 | Llama 3.1 Nemotron Ultra 253B v1 | 61±2 | local | ||
| 221 |  HyperCLOVA X SEED 32B Think | 61±2 | comparison only | ||
| 222 |  Devstral Small 2 24B | 61±3 | local | ||
| more from NVIDIA | |||||
| 98 | Nemotron 3 Ultra 550B-A55B | 69±1 | api $1.35/M +local | ||
| 155 | Nemotron Cascade 2 30B-A3B | 66±1 | local | ||
| 168 | Nemotron 3 Super 120B-A12B | 65±1 | api $0.28/M +local | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #225 of 300 * · 4.1% of its score · 3 tests
Beats o3-mini here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, LFM2.5 1.2B Thinking here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| BFCL v3 | 56.7% ✎ | ▁▁▁▁▁▃▅█▅▄▁▁ beats 64/114 | 4.1% |
| BFCL V4 (December 2025 archive) | 10.0% ✎ | ▁▁▁▂▇▆▆▆█▂▄▂ beats 0/111 | — |
| BFCL v2 Live lab | 74.1% ✎ | █··········█ beats 1/1 | — |
coding▁▁▁▃▆██▇▆▄▃▁ #77 of 458 * · 1.9% of its score · 2 tests
Beats GPT-5.2, Grok 4.20 (0309 reasoning), GPT-5.4 mini here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LiveCodeBench official release_v5 (Nov 1, 2023–Feb 1, 2025) | 74.9% ✎ | █···█····█·█ beats 2/3 | 1.9% |
| LiveCodeBench 2024-08 to 2025-02 (version not stated) lab | 66.3% ✎ | █··········█ beats 1/1 | — |
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #157 of 393 · 30% of its score · 3 tests
Beats MiniMax-M2, MiniCPM5 2B, Granite 4.2 30B here, though they rank higher overall. Trails LFM2.5 1.2B Thinking, LFM2.5 8B-A1B, LFM2.5 2.6B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena (text) | 1347.6 ✎ | ▁▁▁▂▃▃▃▆▆▇█▃ beats 190/379 | 20% |
| Arena-Hard (version/judge unspecified) lab | 87.0% ✎ | ▃▃▂▅▃▆▅█▅▅▂▃ beats 29/30 | 6.4% |
| IFEval (aggregation unspecified) lab | 88.9% ✎ | ▁▁▃▁▂▄▄▆█▃▂▁ beats 102/141 | 3.9% |
math█··········█ #2 of 2 · 34% of its score · 3 tests
other tests2 not in a task · 29% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 76.0% ✎ | ▁▁▁▁▂▄▅▅███▄ beats 176/310 | 16% |
| EQ-Bench 3 (emotional intelligence, judge Elo) | 1139.8 ✎ | ▁·▂▁▄▅▅▃█▄▁▁ beats 17/62 | 12% |
$ where to run it
llama.cpp · recommended
version
- Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.macOS / Linux
curl -LsSf https://llama.app/install.sh | shWindowsirm https://llama.app/install.ps1 | iex - Run Llama 3.1 Nemotron Ultra 253B v1 · needs ~152.3 GB · from huggingface.co/unslothchat in the terminal
llama cli -hf unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF:UD-Q4_K_XL - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF:UD-Q4_K_XL
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Llama 3.1 Nemotron Ultra 253B v1 · needs ~152.3 GB · from huggingface.co/unsloth
- Or download it in the terminal
lms get https://huggingface.co/unsloth/Llama-3_1-Nemotron-Ultra-253B-v1-GGUF@UD-Q4_K_XL -y - Then run lms ls to see its name, and lms chat followed by that name
MLX (Mac)
- Install MLX (Mac) · Apple silicon Macs only, text only.with uv
uv tool install mlx-lmorpip install mlx-lm - Run Llama 3.1 Nemotron Ultra 253B v1 · needs ~174.2 GB, Mac only · from huggingface.co/LibraxisAIchat in the terminal
mlx_lm.chat --model LibraxisAI/Llama-3_1-Nemotron-Ultra-253B-v1-MLX-Q5 - Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model LibraxisAI/Llama-3_1-Nemotron-Ultra-253B-v1-MLX-Q5
Official weights and documentation: huggingface.co/nvidia ↗
$ how the score was worked out · NVIDIA — Llama-Nemotron: Effi… 61%, LMArena leaderboard dataset 20%, EQ-Bench 3 canonical leaderbo… 12% and 2 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The 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's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 NVIDIA — Llama-Nemotron: Efficient Reasoning Models · lab-reported | 6 | 61% |
 LMArena leaderboard dataset | 1 | 20% |
 EQ-Bench 3 canonical leaderboard results | 1 | 12% |
 Berkeley Function Calling Leaderboard result archive | 1 | 4% |
| LiveCodeBench official website results (2025-08-01) | 1 | 2% |
| prior · pull toward the middle | 1% | |
| weighted average | = 61.0 |
not counted (3)
| Berkeley Function Calling Leaderboard result archive | BFCL V4 (December 2025 archive) [correct] | 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 — Llama-Nemotron: Efficient Reasoning Models | BFCL v2 Live [correct] | 74.1% | fewer than three models have this test yet, or this source ran it for this model alone |
| NVIDIA — Llama-Nemotron: Efficient Reasoning Models | LiveCodeBench 2024-08 to 2025-02 (version not stated) [correct] | 66.3% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json