← All comparisons
Magistral Medium 1.2 vs DeepSeek V3.1 Terminus (Reasoning)
Mistral vs DeepSeek — side-by-side benchmark comparison
| Magistral Medium 1.2 | DeepSeek V3.1 Terminus (Reasoning) | |
|---|---|---|
| Intelligence Index | 27.1 | 33.9 |
| Coding Index | 21.7 | 33.7 |
| Math Index | 82.0 | 89.7 |
| Output speed (tok/s) | 39.4 | 0.0 |
| Blended price ($/1M) | $2.75 | $1.91 |
| Time to first token (s) | 0.54s | 0.00s |
| aime | — | — |
| aime 25 | 82.0% | 89.7% |
| artificial analysis coding index | 21.70 | 33.70 |
| artificial analysis intelligence index | 27.10 | 33.90 |
| artificial analysis math index | 82.00 | 89.70 |
| gpqa | 73.9% | 79.2% |
| hle | 9.6% | 15.2% |
| ifbench | 43.0% | 57.0% |
| lcr | 51.3% | 65.0% |
| livecodebench | 75.0% | 79.8% |
| math 500 | — | — |
| mmlu pro | 81.5% | 85.1% |
| scicode | 39.2% | 40.6% |
| tau2 | 52.0% | 37.1% |
| terminalbench hard | 12.9% | 30.3% |
Benchmark data from Artificial Analysis.