← All comparisons
Devstral Small 2 vs GPT-5.4 (Non-reasoning)
Mistral vs OpenAI — side-by-side benchmark comparison
| Devstral Small 2 | GPT-5.4 (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 19.5 | 35.4 |
| Coding Index | 20.7 | 41.0 |
| Math Index | 34.3 | — |
| Output speed (tok/s) | 75.2 | 70.9 |
| Blended price ($/1M) | $0.00 | $5.63 |
| Time to first token (s) | 0.73s | 0.60s |
| aime | — | — |
| aime 25 | 34.3% | — |
| artificial analysis coding index | 20.70 | 41.00 |
| artificial analysis intelligence index | 19.50 | 35.40 |
| artificial analysis math index | 34.30 | — |
| gpqa | 53.2% | 74.8% |
| hle | 3.4% | 10.6% |
| ifbench | 31.2% | 48.4% |
| lcr | 24.0% | 47.3% |
| livecodebench | 34.8% | — |
| math 500 | — | — |
| mmlu pro | 67.8% | — |
| scicode | 28.8% | 47.1% |
| tau2 | 23.4% | 35.1% |
| terminalbench hard | 16.7% | 37.9% |
Benchmark data from Artificial Analysis.