← All comparisons
Hermes 4 - Llama-3.1 405B (Non-reasoning) vs Gemini 2.5 Flash (Reasoning)
Nous Research vs Google — side-by-side benchmark comparison
| Hermes 4 - Llama-3.1 405B (Non-reasoning) | Gemini 2.5 Flash (Reasoning) | |
|---|---|---|
| Intelligence Index | 17.6 | 27.0 |
| Coding Index | 18.1 | 22.2 |
| Math Index | 15.3 | 73.3 |
| Output speed (tok/s) | 40.8 | 205.5 |
| Blended price ($/1M) | $1.50 | $0.85 |
| Time to first token (s) | 0.73s | 10.67s |
| aime | — | 82.3% |
| aime 25 | 15.3% | 73.3% |
| artificial analysis coding index | 18.10 | 22.20 |
| artificial analysis intelligence index | 17.60 | 27.00 |
| artificial analysis math index | 15.30 | 73.30 |
| gpqa | 53.6% | 79.0% |
| hle | 4.2% | 11.1% |
| ifbench | 34.8% | 50.3% |
| lcr | 20.0% | 61.7% |
| livecodebench | 54.6% | 69.5% |
| math 500 | — | 98.1% |
| mmlu pro | 72.9% | 83.2% |
| scicode | 34.6% | 39.4% |
| tau2 | 26.6% | 31.6% |
| terminalbench hard | 9.8% | 13.6% |
Benchmark data from Artificial Analysis.