← All comparisons
Llama 3.1 Nemotron Ultra 253B v1 (Reasoning) vs Hermes 4 - Llama-3.1 70B (Non-reasoning)
NVIDIA vs Nous Research — side-by-side benchmark comparison
| Llama 3.1 Nemotron Ultra 253B v1 (Reasoning) | Hermes 4 - Llama-3.1 70B (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 15.0 | 12.6 |
| Coding Index | 13.1 | 9.2 |
| Math Index | 63.7 | 11.3 |
| Output speed (tok/s) | 52.3 | 94.3 |
| Blended price ($/1M) | $0.90 | $0.20 |
| Time to first token (s) | 0.76s | 0.61s |
| aime | 74.7% | — |
| aime 25 | 63.7% | 11.3% |
| artificial analysis coding index | 13.10 | 9.20 |
| artificial analysis intelligence index | 15.00 | 12.60 |
| artificial analysis math index | 63.70 | 11.30 |
| gpqa | 72.8% | 49.1% |
| hle | 8.1% | 3.6% |
| ifbench | 38.2% | 29.0% |
| lcr | 7.3% | 2.0% |
| livecodebench | 64.1% | 26.9% |
| math 500 | 95.2% | — |
| mmlu pro | 82.5% | 66.4% |
| scicode | 34.7% | 27.7% |
| tau2 | 11.4% | 21.6% |
| terminalbench hard | 2.3% | 0.0% |
Benchmark data from Artificial Analysis.