Mistral Large 3

by Mistral AI · Mistral Large family · released Dec 2, 2025
Open weights Active

Mistral's 675B sparse-MoE flagship (41B active) with native multimodal understanding and 256K context, released under Apache 2.0. One of the largest open-weight models available at launch, also served via Mistral's API.

No pricing recorded for this model yet.

Specifications

Model typeMultimodal
ArchitectureMoE
Parameters675 (41 active)
Context window256K tokens
Max outputNot disclosed tokens
Knowledge cutoffNot disclosed
LicenseApache License 2.0
Modalities Text input Image input Text output
Tags Flagship Long context Vision
End of lifeNot disclosed
WeightsDownload
Last verifiedNot disclosed