models · nvidia
NVIDIA RLHF-aligned refinement of Llama-3.1-70B using HelpSteer2 reward modeling. Tuned for helpfulness and instruction-following.
| Weight file SHA-256 | 32245582399cfe49133a0990681c26a9a4ef2e59c74b4a3f5e1cb7d28f0359bf |
|---|---|
| Method | hf-published-sha256 |
| Publisher | nvidia |
|---|---|
| Origin | United States |
| Licence | LLaMA-3.1 |
| Publisher’s own licence statement | llama3.1 |
| Parameters | 70,553,706,496 (70.6B) |
| Context window | 128K |
| Architecture | llama |
| Layers | 80 |
| Hidden size | 8192 |
| Attention | GQA (64 heads, 8 KV) |
| Native precision | BF16 |
| Pulls recorded here | 96,423 |
| Downloads reported upstream | 11,827 |
| MMLU-Pro | 69 |
|---|---|
| GPQA Diamond | 46.5 |
| AIME 2025 | 11 |
| MATH-500 | 73.3 |
| LiveCodeBench | 16.9 |
| IFBench | 30.7 |
| Humanity's Last Exam | 4.2 |
Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.
Registry record as of 2026-10-14. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.