← All comparisons
Hermes 4 - Llama-3.1 70B (Non-reasoning) vs GPT-5 mini (minimal)
Nous Research vs OpenAI — side-by-side benchmark comparison
| Hermes 4 - Llama-3.1 70B (Non-reasoning) | GPT-5 mini (minimal) | |
|---|---|---|
| Intelligence Index | 12.6 | 20.7 |
| Coding Index | 9.2 | 21.9 |
| Math Index | 11.3 | 46.7 |
| Output speed (tok/s) | 94.3 | 89.3 |
| Blended price ($/1M) | $0.20 | $0.69 |
| Time to first token (s) | 0.61s | 0.73s |
| aime | — | — |
| aime 25 | 11.3% | 46.7% |
| artificial analysis coding index | 9.20 | 21.90 |
| artificial analysis intelligence index | 12.60 | 20.70 |
| artificial analysis math index | 11.30 | 46.70 |
| gpqa | 49.1% | 68.7% |
| hle | 3.6% | 5.0% |
| ifbench | 29.0% | 45.6% |
| lcr | 2.0% | 35.7% |
| livecodebench | 26.9% | 54.5% |
| math 500 | — | — |
| mmlu pro | 66.4% | 77.5% |
| scicode | 27.7% | 36.9% |
| tau2 | 21.6% | 31.9% |
| terminalbench hard | 0.0% | 14.4% |
Benchmark data from Artificial Analysis.