Mistral Large 4
Mistral AI · · 63 days after Shieldstral 1.0 3B
Verified
2 of 2 published values traced to a source1 primary source, 1 archived1 not disclosed by the lab1 not researched yet
Why this status
Newly added. Sources are not archived and no value has been traced to one yet — run `npm run enrich`. Capabilities evidenced from the primary sources: coding: “Human Evaluation We ran an internal evaluation in which expert annotators across coding, computer-aided design (CAD), finance, mathematics and physics compared Mistral Large 4 with” | agentic: “Mistral Large 4 New Open-weight hybrid instruct-and-reasoning MoE with multimodal input; unifies instruction, reasoning, and agentic capabilities in a single model, state-of-the-ar”.
What these states mean · How records are judged · Has this record changed?
A natively multimodal mixture-of-experts model with 1 trillion total and 52 billion active parameters, released as a public preview API on Mistral Studio. Mistral said it would release the weights by the end of the month; none were public at announcement.
Details
| Released | October 6, 2026stated in mistral.aiMistral Large 4 was released on October 6, 2026.
|
|---|---|
| Company | Mistral AI |
| Family | Mistral |
| Lineage | after Mistral Large 3 |
| Era | Agentic systems |
| Type | Language model · LLM |
| Modalities | Not recorded |
| Capabilities | Coding, Agentic |
| Weights | Proprietary |
| Licence | Not applicable — proprietary |
| Context window | Not disclosed |
| Parameters | 1Tstated in mistral.aiMistral Large 4 has 1,000,000,000,000 parameters.
|
Benchmarks
Each row is a dated claim by a named party, not a property of the model. A later revision adds a row rather than replacing one. This project publishes no composite or overall score.
| Benchmark | Score | Reported by | Date |
|---|---|---|---|
| Terminal-Bench | 28.3 | Vendor reported | October 6, 2026 |
Current information, elsewhere
This record is a historical one: every figure above is traced to a dated primary source. For what is true today — live pricing, current rankings, availability — these are better maintained than anything here could be.
- Current price and providers OpenRouter ↗
- Current benchmark scores Artificial Analysis ↗
Sources
Record status: verified · confidence 90/100 · 1 of 1 primary
Newly added. Sources are not archived and no value has been traced to one yet — run `npm run enrich`. Capabilities evidenced from the primary sources: coding: “Human Evaluation We ran an internal evaluation in which expert annotators across coding, computer-aided design (CAD), finance, mathematics and physics compared Mistral Large 4 with” | agentic: “Mistral Large 4 New Open-weight hybrid instruct-and-reasoning MoE with multimodal input; unifies instruction, reasoning, and agentic capabilities in a single model, state-of-the-ar”.
- mistral.ai Official announcement archived
Primary means published by the organisation that made the model. Verified records require at least one.
Mistral family
- Mistral 7B September 27, 2023
- Mistral Large 2 July 24, 2024
- Mistral Small 3 January 30, 2025
- Mistral Large 3 December 2, 2025
- Ministral 3 3B December 2, 2025
- Ministral 3 8B December 2, 2025
- Ministral 3 14B December 2, 2025
- Mistral Small 4 March 16, 2026
- Mistral Medium 3.5 May 22, 2026
- Shieldstral 1.0 3B August 4, 2026
- Mistral Large 4 October 6, 2026
What changed from Mistral Large 3
Compared with the previous release in this series. A field appears only when both records state a value — where one does not, it is listed as uncomparable rather than dropped, because a gap in our research is not a finding about the model.
- Parameters675B → 1T
- WeightsOpen → Proprietary
Not comparable: Context window (not disclosed for both releases); Capabilities and modalities (not researched on both releases)
Compare Mistral Large 4 with another model →
See Mistral Large 4 on the interactive timeline →