0

Mistral Large 3

Open weights

Mistral AI's December 2025 flagship - a 675B-total / 41B-active open-weights MoE under Apache 2.0.

Publisher
Mistral AI
Params
675B total / 41B active
Context
256K
Input
Output
License
Apache-2.0
Released
Dec 2025

Cite

Notes

Only stored in your browser.

Attribution

Benchmark scores
AA
Attribution policy →
Context
256K
Input
$0.50
/1M tok
Output
$1.50
/1M tok
Blended
$0.75
#282/ 414/1M · 3:1
Speed
63
#153/ 414tok/s

Price & speed via Artificial Analysis · ranked across tracked models

Reported on 9 evals across 6 domains - best 80.7% on MMLU-Pro · #67 of 342

Reported eval scores

9