← All comparisons
Llama 3.3 Instruct 70B vs Hermes 4 - Llama-3.1 70B (Reasoning)
Meta vs Nous Research — side-by-side benchmark comparison
| Llama 3.3 Instruct 70B | Hermes 4 - Llama-3.1 70B (Reasoning) | |
|---|---|---|
| Intelligence Index | 14.5 | 16.0 |
| Coding Index | 10.7 | 14.4 |
| Math Index | 7.7 | 68.7 |
| Output speed (tok/s) | 88.1 | 92.8 |
| Blended price ($/1M) | $0.62 | $0.20 |
| Time to first token (s) | 0.59s | 0.64s |
| aime | 30.0% | — |
| aime 25 | 7.7% | 68.7% |
| artificial analysis coding index | 10.70 | 14.40 |
| artificial analysis intelligence index | 14.50 | 16.00 |
| artificial analysis math index | 7.70 | 68.70 |
| gpqa | 49.8% | 69.9% |
| hle | 4.0% | 7.9% |
| ifbench | 47.1% | 31.3% |
| lcr | 15.0% | 6.7% |
| livecodebench | 28.8% | 65.3% |
| math 500 | 77.3% | — |
| mmlu pro | 71.3% | 81.1% |
| scicode | 26.0% | 34.1% |
| tau2 | 26.6% | 22.5% |
| terminalbench hard | 3.0% | 4.5% |
Benchmark data from Artificial Analysis.