AI KNOWLEDGE DESK

Models · entities · concepts · comparisons · practical tools

GETLLMS.ORG
← All entities
Multimodal reasoning modelsChecked

Mistral Large 4

Mistral Large 4 is Mistral’s October 6, 2026 public-preview multimodal model, callable as mistral-large-4 with a 1M context window. API preview access is available; the announced open-weight checkpoint is still upcoming.

Why it matters

An API preview and a downloadable checkpoint are different deployment choices. Developers can trial the hosted API now while keeping self-hosting, final weight licensing and production migration contingent on the actual weight release.

API preview and architecture

The official card lists a granular Mixture-of-Experts model with 1.05T total parameters, 49B active per token and a 1.6B vision encoder. The upcoming checkpoint name includes 52B, counting embedding and output layers; that is not a contradictory active-per-token number. The API supports structured outputs, function calling, Document QnA, batch and agent/conversation surfaces.

Introductory and base token rates

Mistral lists a two-week 50% launch discount: $0.68 input, $0.07 cached input and $2.09 output per million tokens. Undiscounted listed rates are $1.36, $0.14 and $4.18 respectively. Check the live pricing page at request time rather than assuming the discount is permanent.

Weights and license remain a separate gate

The official checkpoint is mistralai/Mistral-Large-4.0-1T05-A52B, currently marked upcoming. Mistral says weights are planned for the end of October. Its final weight license is not confirmed here; an Open label or licenses used by other Mistral models do not determine this release’s redistribution terms.

How to assess the preview

Start with representative tasks and matched tool, effort and latency budgets. Mistral links an official CTF speedrun demonstration, which is not a general benchmark or proof of safe production behavior. Validate the API contract and rollout terms before replacing an existing model.

Useful first jobs

  • Trial a multimodal API for document and coding tasks.
  • Evaluate structured outputs and tool use with a defined harness.
  • Plan self-hosting only after weights and final terms are released.

Frequently asked questions

Can I download Mistral Large 4 weights today?

The checked October 6 official release marks the weights as upcoming, planned for the end of October. API preview availability does not prove a downloadable checkpoint exists.

Is the launch price permanent?

No. Mistral’s changelog describes two weeks of 50% launch pricing; check the current rate before budgeting.

Sources and evidence

Mistral’s model card and changelog control API identity, preview status, context and prices. Its upcoming Hugging Face card does not establish released weights. Performance and CTF demonstrations are first-party examples, not independent evaluation.

Continue reading