← All comparisons
Claude 3.5 Sonnet (June '24) vs Devstral Small (Jul '25)
Anthropic vs Mistral — side-by-side benchmark comparison
| Claude 3.5 Sonnet (June '24) | Devstral Small (Jul '25) | |
|---|---|---|
| Intelligence Index | 14.2 | 15.2 |
| Coding Index | 26.0 | 12.1 |
| Math Index | — | 29.3 |
| Output speed (tok/s) | 0.0 | 183.4 |
| Blended price ($/1M) | $6.56 | $0.15 |
| Time to first token (s) | 0.00s | 0.40s |
| aime | 9.7% | 0.3% |
| aime 25 | — | 29.3% |
| artificial analysis coding index | 26.00 | 12.10 |
| artificial analysis intelligence index | 14.20 | 15.20 |
| artificial analysis math index | — | 29.30 |
| gpqa | 56.0% | 41.4% |
| hle | 3.7% | 3.7% |
| ifbench | — | 34.6% |
| lcr | — | 17.0% |
| livecodebench | — | 25.4% |
| math 500 | 69.5% | 63.5% |
| mmlu pro | 75.1% | 62.2% |
| scicode | 31.6% | 24.3% |
| tau2 | — | 28.4% |
| terminalbench hard | — | 6.1% |
Benchmark data from Artificial Analysis.