NVIDIA model product
A 253B parameter derivative of Meta Llama 3.1 405B Instruct, developed by NVIDIA using Neural Architecture Search (NAS) and vertical compression. It underwent multi-phase post-training (SFT for Math, Code, Reasoning, Chat, Tool Calling; RL with GRPO) to enhance reasoning and instruction-following. Optimized for accuracy/efficiency tradeoff on NVIDIA GPUs. Supports 128k context.
Updated Aug 10, 2026. Default version: Llama 3.1 Nemotron Ultra 253B v1
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Preference and agent-evaluation signals from published Arena datasets.
| text style control | math | 162 | 1380.4 | 209 | N/A | |
| text style control | hard prompts english | 173 | 1397.3 | 448 | N/A | |
| text | math | 177 | 1359.5 | 209 | N/A | |
| text style control | creative writing | 177 | 1331.9 | 444 | N/A | |
| text style control | instruction following | 177 | 1351.3 | 660 | N/A | |
| text style control | hard prompts | 181 | 1374.8 | 713 | N/A | |
| text style control | english | 182 | 1375.6 | 1,491 | N/A | |
| text | creative writing | 183 | 1314.6 | 444 | N/A | |
| text style control | industry software and it services | 187 | 1387.4 | 588 | N/A | |
| text style control | coding | 190 | 1391.3 | 367 | N/A | |
| text style control | multi turn | 193 | 1342.2 | 374 | N/A | |
| text style control | overall | 194 | 1347.6 | 2,549 | N/A | |
| text | instruction following | 198 | 1308.2 | 660 | N/A | |
| text style control | exclude ties | 198 | 1304.1 | 1,699 | N/A | |
| text style control | industry business and management and financial operations | 199 | 1342.1 | 254 | N/A | |
| text style control | industry life and physical and social science | 200 | 1360.4 | 498 | N/A | |
| text style control | industry writing and literature and language | 200 | 1320.4 | 712 | N/A | |
| text | english | 201 | 1344.5 | 1,491 | N/A | |
| text | industry writing and literature and language | 202 | 1300.3 | 712 | N/A | |
| text style control | industry entertainment and sports and media | 202 | 1303.0 | 532 | N/A | |
| text | hard prompts english | 203 | 1336.2 | 448 | N/A | |
| text | multi turn | 203 | 1317.0 | 374 | N/A | |
| text | overall | 205 | 1319.6 | 2,549 | N/A | |
| text | hard prompts | 205 | 1316.1 | 713 | N/A | |
| text | industry life and physical and social science | 205 | 1323.9 | 498 | N/A | |
| text style control | longer query | 206 | 1341.5 | 321 | N/A | |
| text | exclude ties | 207 | 1263.8 | 1,699 | N/A | |
| text | industry business and management and financial operations | 208 | 1301.1 | 254 | N/A | |
| text | industry software and it services | 208 | 1327.5 | 588 | N/A | |
| text | industry entertainment and sports and media | 209 | 1276.9 | 532 | N/A | |
| text | non english | 213 | 1282.8 | 1,058 | N/A | |
| text style control | russian | 213 | 1313.5 | 303 | N/A | |
| text | russian | 214 | 1283.4 | 303 | N/A | |
| text | coding | 217 | 1312.4 | 367 | N/A | |
| text style control | non english | 218 | 1306.1 | 1,058 | N/A | |
| text | longer query | 219 | 1298.8 | 321 | N/A |
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| Cortecs | llama-3.1-nemotron-ultra-253b-v1 | global | $0.598 | $1.79 | 128K | |
| Nebius Token Factory | nvidia/Llama-3_1-Nemotron-Ultra-253B-v1 | global | $0.60 | $1.8 | 128K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Llama 3.1 Nemotron Ultra 253B v1 | 62.6 | 253B | N/A | N/A | No | Llama 3.1 Community License |
A concise description based on the published model registry.
A 253B parameter derivative of Meta Llama 3.1 405B Instruct, developed by NVIDIA using Neural Architecture Search (NAS) and vertical compression. It underwent multi-phase post-training (SFT for Math, Code, Reasoning, Chat, Tool Calling; RL with GRPO) to enhance reasoning and instruction-following. Optimized for accuracy/efficiency tradeoff on NVIDIA GPUs. Supports 128k context.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Llama 3.1 Nemotron Ultra 253B.
Llama 3.1 Nemotron Ultra 253B's default version was released on Apr 7, 2025.
No official standard PAYG price is currently available for Llama 3.1 Nemotron Ultra 253B. The lowest tracked third-party offer starts at $0.598 input and $1.79 output via Cortecs.
Llama 3.1 Nemotron Ultra 253B is published under NVIDIA in the model registry.
The current registry does not publish a context window for the default version.
No. The default version is not marked as having publicly available weights.
2 published provider offerings are linked to the default version.
Nearby ranked alternatives include MAI Thinking 1, o1, Llama 3.3 Nemotron Super 49B.