Mistral · model
Mistral Large 4 Preview benchmark scores
As of 2026-10-07, Mistral Large 4 Preview (Mistral) has measured scores on 7 benchmarks and estimated scores on 7 more.
| Benchmark | Score | Source |
|---|---|---|
| Terminal-Bench 2.1 | 86.9% | estimated ± 4.0 pp, high confidence |
| Terminal-Bench 4.0 | 26.8% | measured |
| Terminal-Bench Science 0.1 | 4.1% | estimated ± 6.5 pp, medium confidence |
| GPQA Diamond | 90.0% | estimated ± 1.3 pp, high confidence |
| SciCode | 54.2% | measured |
| CritPt | 10.6% | measured |
| GDP.pdf | 18.6% | measured |
| τ³-Bench Banking | 47.4% | estimated ± 5.0 pp, medium confidence |
| MMMU-Pro | 76.0% | measured |
| AA-AnalystAgent | 47.8% | estimated ± 7.8 pp, medium confidence |
| Harvey LAB | 92.8% | estimated ± 3.2 pp, high confidence |
| Humanity's Last Exam | 35.0% | measured |
| AutomationBench | 59.9% | measured |
| EnterpriseOps-Gym | 42.3% | estimated ± 6.2 pp, low confidence |