Mistral Large 4
API ID mistral-large-4
Launched October 6, 2026
Mistral Large 4 is a preview Mistral model, released October 2026. It has a 1,000,000-token context window. It takes image and PDF input and supports function calling, structured outputs and reasoning. Standard price: $0.68 input and $2.09 output per 1M tokens.
Mistral Large 4 is a state-of-the-art, open-weight, general-purpose multimodal model with a granular Mixture-of-Experts architecture. It features 52B active parameters and 1.05T total parameters, and a 1.6B vision encoder.
- Status
- Active
- Released
- October 2026
- Shutdown
- none announced
- Replacement
- —
No deprecation or shutdown date has been announced as of October 15, 2026.
Benchmarks
Scores as each source published themComparing with Mistral Large 3, the previous model in this line. A score is compared only with the same benchmark from the same source, at the same setting where both have it. Full comparison: prices, limits, features
Reported by Mistral
| Benchmark | Mistral Large 4 | Mistral Large 3 | Change | Setting | Source |
|---|---|---|---|---|---|
| Cybench | 93.0% | — | post | ||
| DeepSWE v1.1 | 61.7% | — | post | ||
| SWE-Atlas-QnA | 59.4% | — | post | ||
| Terminal-Bench 4 | 28.3% | — | post |
Run by Mistral with its own settings and test setup, so not directly comparable with other companies' numbers. Each line links to the post (Oct 6, 2026).
Arena
| Arena | Mistral Large 4 | Mistral Large 3 | Change | Rank | Date | Source |
|---|---|---|---|---|---|---|
TextBy category
| 1431 ±8 | 1414 ±3 | +17 | 112 of 415 6,607 votes | Oct 13, 2026 | source |
| Text: Coding | 1499 ±15 | 1471 ±5 | +28 | 82 of 410 1,489 votes | Oct 13, 2026 | source |
| WebDev | 1542 ±14 | 1230 ±25 | +312 | 42 of 142 2,194 votes | Oct 13, 2026 | source |
| Agent | -0.032 ±0.017 | not rated | 39 of 54 14,022 sessions | Oct 14, 2026 | source |
Rated as mistral-large-4; Mistral Large 3 as mistral-large-3. Text and WebDev ratings come from people voting between two anonymous answers; Text is style-controlled, as on arena.ai. Agent is Arena's separate score for real agent sessions, on its own scale. Not rated on Vision, where Mistral Large 3 is. Data: Arena (arena.ai), CC BY 4.0.
Epoch AI
| Benchmark | Mistral Large 4 | Mistral Large 3 | Change | Setting | Run | Source |
|---|---|---|---|---|---|---|
| GPQA Diamond | 88.1% | not run | high effort | Oct 12, 2026 | source | |
| OTIS Mock AIME 2024-2025 | 98.3% | not run | high effort | Oct 12, 2026 | source | |
| SimpleQA Verified | 20.0% | not run | high effort | Oct 9, 2026 | source |
Run independently by Epoch AI. Data: Epoch AI, Capabilities & benchmarking (epoch.ai/benchmarks), CC BY.
Features
Compare models| Feature | Value | Since | Source |
|---|---|---|---|
| Structured outputsThe model can be held to a JSON schema you supply, so every reply parses without retries or cleanup. | Supported | in place when tracking began (October 6, 2026) | source |
| Function callingAlso called tool use. You describe functions; the model replies with a function name and arguments for your code to run, then uses the result. | Supported | in place when tracking began (October 6, 2026) | source |
| Batch APISend many requests as one job that runs in the background, usually finished within 24 hours, at a lower price than live requests. | Supported | in place when tracking began (October 6, 2026) | source |
| Image inputImages such as screenshots, photos and charts can go in the prompt for the model to read. | Supported | in place when tracking began (October 6, 2026) | source |
| PDF inputPDF files can go in the prompt directly, without converting them to text first. | Supported | in place when tracking began (October 6, 2026) | source |
| Reasoning / thinkingThe model can work through a problem step by step before it answers: usually better answers to hard questions, at the cost of more tokens and time. | Supported | first seen October 7, 2026 | source |
| Context windowThe most the model can work with in one request, in tokens: your prompt, conversation history and files. A token is about three-quarters of an English word. | 1,000,000 tokens | in place when tracking began (October 6, 2026) | source |
Not published by Mistral: Prompt caching, Fine-tuning, Built-in web search, temperature parameter, top_p parameter, logprobs, Max output tokens.
Price history
USD per 1M tokens, standard tierWe started recording this model's prices on Oct 6, 2026. The history grows with each daily check.
Show as a table
| From | Type | Price |
|---|---|---|
| Oct 6, 2026 | Input | $0.68 |
| Oct 6, 2026 | Output | $2.09 |
Prices
USD per 1M tokens, every type and band, by tier| Type | Price $/1M | In effect |
|---|---|---|
| Standard | ||
| input | $0.68 | |
| cached input | $0.07 | |
| output | $2.09 | |
Changelog
Changes we detected on Mistral's pages- Feature listed
mistral-large-4: Reasoning / thinking supported
Info
- Added
Mistral added mistral-large-4
Info
Related news
- ReleasesInfoMistral added mistral-large-4
- InfoMistral Large 4
- ReleasesInfoIntroducing Mistral Large 4
Email alerts
Get an email if mistral-large-4 changes
We'll email you the same morning we spot a deprecation notice, a new or moved shutdown date, a price change, or the model leaving Mistral's model list.
No password: one email to confirm, and you can unsubscribe any time. Privacy
Last checked · Sources: Mistral models