← All comparisons
Claude 4.5 Haiku (Reasoning) vs Devstral Small (Jul '25)
Anthropic vs Mistral — side-by-side benchmark comparison
| Claude 4.5 Haiku (Reasoning) | Devstral Small (Jul '25) | |
|---|---|---|
| Intelligence Index | 37.1 | 15.2 |
| Coding Index | 32.6 | 12.1 |
| Math Index | 83.7 | 29.3 |
| Output speed (tok/s) | 142.2 | 183.4 |
| Blended price ($/1M) | $2.19 | $0.15 |
| Time to first token (s) | 10.48s | 0.40s |
| aime | — | 0.3% |
| aime 25 | 83.7% | 29.3% |
| artificial analysis coding index | 32.60 | 12.10 |
| artificial analysis intelligence index | 37.10 | 15.20 |
| artificial analysis math index | 83.70 | 29.30 |
| gpqa | 67.2% | 41.4% |
| hle | 9.7% | 3.7% |
| ifbench | 54.3% | 34.6% |
| lcr | 70.3% | 17.0% |
| livecodebench | 61.5% | 25.4% |
| math 500 | — | 63.5% |
| mmlu pro | 76.0% | 62.2% |
| scicode | 43.3% | 24.3% |
| tau2 | 54.7% | 28.4% |
| terminalbench hard | 27.3% | 6.1% |
Benchmark data from Artificial Analysis.