Mistral AI
Mistral Large 4
Mistral's flagship preview for coding and professional work. A sparse 1.05T-parameter model activates 49B parameters per token.
Replaces Mistral Large 3: input $0.525 → $0.714 per 1M (+36%), output $1.575 → $2.1945 per 1M (+39%), context 262K → 524K, max output 256K → 262K.
Run this model
mistral/mistral-large-4
Use an endpoint supported by this model.
Get a key for this setup →const response = await fetch('https://api.minirouter.sh/v1/chat/completions', { method: 'POST', headers: { Authorization: `Bearer ${process.env.MINIROUTER_KEY}`, 'Content-Type': 'application/json', }, body: JSON.stringify({ model: 'mistral/mistral-large-4', messages: [{ role: 'user', content: 'Why is the sky blue?' }], max_tokens: 1024, }),}) const data = await response.json()console.log(data.choices[0].message.content)MiniRouter currently supports text input and text output for this model. Use an explicit max_tokens limit. Mistral's native vision support and 1M context do not describe this route; its current limits appear below.
Model overview
- Input / 1M
- $0.714
- All-in MiniRouter rate
- Output / 1M
- $2.1945
- All-in MiniRouter rate
- Context
- 524K
- 524,288 tokens
- Max output
- 262K
- tokens
- Architecture
- 1.05T MoE
- 49B active parameters
- Stage
- Preview
- Training is still underway
- Announced
- 6 Oct 2026
- publisher release
Coding scores from the preview.
Mistral reports these results for Large 4. Each benchmark measures a different task; the scores are not MiniRouter measurements or a composite ranking.
Mistral-reported
DeepSWE v1.1
Public-preview result. Consult the publisher's evaluation details before comparing models.
Mistral Large 4
0 to 100%
- Mistral Large 461.7%
View chart data
| Model | Reported result |
|---|---|
| Mistral Large 4 | 61.7% |
Mistral-reported
Terminal-Bench 4.0
Public-preview result. Consult the publisher's evaluation details before comparing models.
Mistral Large 4
0 to 100%
- Mistral Large 428.3%
View chart data
| Model | Reported result |
|---|---|
| Mistral Large 4 | 28.3% |
Start with code and long documents.
- [ 01 ]
Code changes
Try a bounded implementation task with existing tests. Compare completed-task quality and cost, including retries.
- [ 02 ]
Document analysis
Supply text and ask for citations to the supplied material. Keep input and generated output within the route's shared context window.
- [ 03 ]
Specialist review
Use domain experts to verify conclusions. Publisher benchmark results do not establish accuracy on your data.
Use the route's limits.
- [ 01 ]
Text only here
Do not send images to this route. The model's native visual grounding capability is separate from MiniRouter's admitted input types.
- [ 02 ]
Preview prices
The release article records the provider's introductory rates. Current MiniRouter prices above include its markup; no future discount or cache hit is guaranteed.
- [ 03 ]
Weights are planned
Mistral plans to release weights by the end of October. Until then, the model is available through the API preview.
One route, standard clients.
https://api.minirouter.sh/v1mistral/mistral-large-4/v1/chat/completions/v1/messages/v1/responsesConfigured providers. Priority follows the live catalog order.
| Provider | Priority | Input / 1M | Output / 1M | Parameters | Access |
|---|---|---|---|---|---|
vercel | 1 | $0.714 | $2.1945 | max_tokens, temperature, stop, tools, tool_choice, reasoning, include_reasoning | MiniRouter balance |