← All comparisons
Gemma 4 31B (Non-reasoning) vs Hermes 4 - Llama-3.1 405B (Reasoning)
Google vs Nous Research — side-by-side benchmark comparison
| Gemma 4 31B (Non-reasoning) | Hermes 4 - Llama-3.1 405B (Reasoning) | |
|---|---|---|
| Intelligence Index | 32.3 | 18.6 |
| Coding Index | 33.9 | 16.0 |
| Math Index | — | 69.7 |
| Output speed (tok/s) | 18.0 | 38.6 |
| Blended price ($/1M) | $0.20 | $1.50 |
| Time to first token (s) | 0.60s | 0.79s |
| aime | — | — |
| aime 25 | — | 69.7% |
| artificial analysis coding index | 33.90 | 16.00 |
| artificial analysis intelligence index | 32.30 | 18.60 |
| artificial analysis math index | — | 69.70 |
| gpqa | 76.3% | 72.7% |
| hle | 11.5% | 10.3% |
| ifbench | 53.5% | 32.7% |
| lcr | 36.0% | 20.7% |
| livecodebench | — | 68.6% |
| math 500 | — | — |
| mmlu pro | — | 82.9% |
| scicode | 41.1% | 25.2% |
| tau2 | 65.5% | 22.2% |
| terminalbench hard | 30.3% | 11.4% |
Benchmark data from Artificial Analysis.