← All comparisons
Hermes 4 - Llama-3.1 70B (Reasoning) vs Qwen3 235B A22B 2507 (Reasoning)
Nous Research vs Alibaba — side-by-side benchmark comparison
| Hermes 4 - Llama-3.1 70B (Reasoning) | Qwen3 235B A22B 2507 (Reasoning) | |
|---|---|---|
| Intelligence Index | 16.0 | 29.5 |
| Coding Index | 14.4 | 23.2 |
| Math Index | 68.7 | 91.0 |
| Output speed (tok/s) | 92.8 | 62.5 |
| Blended price ($/1M) | $0.20 | $0.84 |
| Time to first token (s) | 0.64s | 1.21s |
| aime | — | 94.0% |
| aime 25 | 68.7% | 91.0% |
| artificial analysis coding index | 14.40 | 23.20 |
| artificial analysis intelligence index | 16.00 | 29.50 |
| artificial analysis math index | 68.70 | 91.00 |
| gpqa | 69.9% | 79.0% |
| hle | 7.9% | 15.0% |
| ifbench | 31.3% | 51.2% |
| lcr | 6.7% | 67.0% |
| livecodebench | 65.3% | 78.8% |
| math 500 | — | 98.4% |
| mmlu pro | 81.1% | 84.3% |
| scicode | 34.1% | 42.4% |
| tau2 | 22.5% | 53.2% |
| terminalbench hard | 4.5% | 13.6% |
Benchmark data from Artificial Analysis.