The largest Mistral model, a natively multimodal 1-trillion-parameter MoE with 49B active. Launched as a public API preview while it is red-teamed with cybersecurity partners and state authorities; Mistral says open weights follow by the end of October.
SI v3 uses weighted percentages from one Artificial Analysis configuration. Ranking requires all five benchmarks. Partial scores average only available weights. Scores are not comparable to SI v2 or a probability of reaching singularity.
Compared with 56 models released within six months of it, across 3 benchmarks where scores spread out.
| Benchmark | Score | Rank |
|---|---|---|
HLEArtificial Analysis Challenging multidisciplinary questions evaluated by Artificial Analysis | 35% | #51 / 96 |
| Benchmark / source | Score |
|---|---|
HLE Artificial AnalysisMistral Large 4 Preview | 35% |
MMMU-Pro Artificial AnalysisMistral Large 4 Preview | 76.4% |
Terminal-Bench 4.0 Artificial AnalysisMistral Large 4 Preview | 26.8% |
SciCode Artificial AnalysisMistral Large 4 Preview | 54.2% |
AA-LCR Artificial AnalysisMistral Large 4 Preview | 81.3% |
CritPt Artificial AnalysisMistral Large 4 Preview | 10.6% |