Mistral Large 3
Mistral AI class A live
Mistral's most-capable model: sparse MoE (41B active / 675B total), Apache-2.0 open weights, 256K context, multimodal, multilingual.
Layer 1 — factual axes
Layer 2 — output quality
16 / 100
Mistral’s most-capable model: sparse MoE (41B active / 675B total), Apache-2.0 open weights, 256K context, multimodal, multilingual.