Mistral opens the Large 4 multimodal model preview

The short version

Mistral has opened an API preview for Mistral Large 4, a multimodal MoE model with 1.05T total parameters and 49B active parameters. Its model page lists a 1M-token context window; weights are planned for release later this month and are not yet available.

What changed

Mistral added a Large 4 preview API in Studio while its open-weight release remains planned rather than delivered.

What it means for you

Teams needing multimodal long context or Agent APIs can test it now, but must not treat the unreleased weights as a current deployment option.

Mistral opened an API preview for Mistral Large 4 on October 6. The company describes it as a general-purpose multimodal MoE model available to try in Mistral Studio; its public weights are planned for the end of the month, so it is not yet a delivered self-hosting option.

The official model page lists 1.05T total parameters, 49B active parameters, a 1.6B vision encoder, and a 1M-token context window. The preview exposes chat completions, Conversations, function calling, structured outputs, document Q&A, batching, and Agents APIs; check the live documentation for pricing and availability.

This gives teams evaluating coding agents, document understanding, or visual workflows a new model to test today. Preview status and unreleased weights mean deployment decisions should still include a fallback. See the Mistral announcement and model documentation.

Mistral Large 4

Fact check

VerifiedMistral opened a public API preview for Mistral Large 4 on October 6, 2026, and says its weights will be released by the end of the month.Evidence

VerifiedMistral documents 1.05T total parameters, 49B active parameters, a 1M-token context window, structured outputs, function calling, and Agents APIs for Large 4.Evidence