← All comparisons
Gemini 3.1 Pro Preview vs Hermes 4 - Llama-3.1 405B (Non-reasoning)
Google vs Nous Research — side-by-side benchmark comparison
| Gemini 3.1 Pro Preview | Hermes 4 - Llama-3.1 405B (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 57.2 | 17.6 |
| Coding Index | 55.5 | 18.1 |
| Math Index | — | 15.3 |
| Output speed (tok/s) | 136.8 | 40.8 |
| Blended price ($/1M) | $4.50 | $1.50 |
| Time to first token (s) | 25.68s | 0.73s |
| aime | — | — |
| aime 25 | — | 15.3% |
| artificial analysis coding index | 55.50 | 18.10 |
| artificial analysis intelligence index | 57.20 | 17.60 |
| artificial analysis math index | — | 15.30 |
| gpqa | 94.1% | 53.6% |
| hle | 44.7% | 4.2% |
| ifbench | 77.1% | 34.8% |
| lcr | 72.7% | 20.0% |
| livecodebench | — | 54.6% |
| math 500 | — | — |
| mmlu pro | — | 72.9% |
| scicode | 58.9% | 34.6% |
| tau2 | 95.6% | 26.6% |
| terminalbench hard | 53.8% | 9.8% |
Benchmark data from Artificial Analysis.