Mistral · Mistral 4
Mistral Large 4
Mistral's first trillion-parameter MoE (about 1.05T total, 52B active) with image input, in public preview. AA rates it the strongest model from outside the US and China.
- AA Intelligence
- 38#34 of 151
- Input / output per 1M
- $1.36/ $4.18
- Output speed
- —
- Context window
- 1M tokens
Composite indexes
- Artificial Analysis Intelligence Index#34 of 15138
Leader: Claude Opus 5.5 at 58
Coding
- Terminal-Bench 4.0#23 of 9628.3%27%
Leader: Claude Sonnet 5.5 at 70.6%
- SciCode#30 of 10054%
Leader: Claude Opus 5.5 at 66.9%
Agents and tool use
- GDPval-AA#32 of 1091422
Leader: Muse Wonder Full at 1857
Reasoning and knowledge
- Humanity’s Last Exam#52 of 15335%
Leader: Claude Sonnet 5.5 at 64.5%
- CritPt#41 of 14411%
Leader: Muse Wonder Full at 33%
Vision and long context
- AA Long Context Reasoning#27 of 14481%
Leader: Kimi K3 at 88.7%
Human preference
- LMArena Text#46 of 571429
Leader: Gemini 4 Argon at 1525
- LMArena WebDev#32 of 581541
Leader: Claude Opus 5.5 at 1818
Other published results
| Result | Score | Reported by | Source |
|---|---|---|---|
| AutomationBench-AA (partial score) | 60% | independent | Artificial Analysis: Mistral Large 4 ↗ |
| AA-Omniscience Index | -5 index (-100 to 100) | independent | Artificial Analysis: Mistral Large 4 ↗ |
| DeepSWE v1.1 | 61.7% | vendor | Mistral: Introducing Mistral Large 4 ↗ |
| AA Coding Agent Index | 49.8 index | vendor | Mistral: Introducing Mistral Large 4 ↗ |