Mistral AI · checked October 13, 2026
How do I run Mistral Large 4 in OpenCode, Claude Code or Codex?
Mistral Large 4 runs on Standard Compute at https://api.stdcmpt.com/v1 with the model ID mistralai/mistral-large-4-0. The same key works in OpenCode, Claude Code and Codex.
| Model | Mistral's open-weight flagship, released October 6, 2026 as a public preview. Mistral describes a mixture-of-experts model with 52B active and 1.05T total parameters, a 1.6B vision encoder and a 1M-token context window, built for reasoning, coding and agent work. |
|---|---|
| Model ID | mistralai/mistral-large-4-0Mistral's own API calls it mistral-large-4 |
| Base URL | https://api.stdcmpt.com/v1Claude Code uses https://api.stdcmpt.com |
| Context | 1,048,576 tokens; about 850K input and 50K output per request |
| Input | Text and images |
| Reasoning | No setting on the Models page; it runs at the model's default |
Copy a working config
Paste the config for your agent. The same ID works in all three, and Request Logs show which model answered each call.
{
"$schema": "https://opencode.ai/config.json",
"model": "standardcompute/mistralai/mistral-large-4-0",
"provider": {
"standardcompute": {
"npm": "@ai-sdk/openai-compatible",
"name": "Standard Compute",
"options": {
"baseURL": "https://api.stdcmpt.com/v1",
"apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
},
"models": {
"mistralai/mistral-large-4-0": { "name": "Mistral Large 4" },
"standardcompute": { "name": "Smart routing" }
}
}
}
}
Replace YOUR_API_KEY with a key from the dashboard; OpenCode reads it from STANDARDCOMPUTE_API_KEY. Full setup: OpenCode, Claude Code, Codex CLI.
What it costs
| Per 1M tokens | List price | Mistral sale price |
|---|---|---|
| Input | $1.36 | $0.68 |
| Cached input | $0.14 | $0.07 |
| Output | $4.18 | $2.09 |
Mistral API rates. Mistral shows the sale price as 50% off and gives no end date. Batch requests are half price. Source: Mistral API pricing, checked October 13, 2026.
Averages from coding-agent traffic on our gateway, Sep 13 to Oct 12, 2026: 87% of prompt tokens are cache hits and a turn returns about 700 tokens. Vendor list prices. At Mistral's current sale price the same turn costs $0.015.
The $19 Starter plan includes a $20 compute budget, roughly 670 turns like this one at list price. Large 4 also fits as the European rung in a mixed selection: select it next to a frontier model and routing decides which one answers each request.
What works differently here
- Mistral is selling Large 4 at half price with no end date. While that lasts, paying Mistral directly for a Large 4-only workload costs half of the list prices above.
- Mistral still labels Large 4 a public preview, so its behavior and price can change. Artificial Analysis scores the preview 38 on its Intelligence Index and 26.8% in its Terminal-Bench 4.0 run, below the current Claude, GPT and Gemini flagships.
- There is no reasoning control for it on the Models page yet, and requests stop at about 850K input tokens, short of the full 1M window.
Other ways to run Mistral Large 4
| Mistral API $0.68 / $2.09 per 1M | First-party, at a 50% sale price from the $1.36 / $4.18 list rate. Batch is half price. |
|---|---|
| OpenRouter Pay per token | Served by Mistral at its sale price, with one account for other vendors. |
| Self-host GPU + power | Mistral lists the weights as open. At 1.05T total parameters it needs multi-GPU server hardware. |
| Standard Compute Monthly plan | Large 4 next to Claude, GPT and other open models on one key, with image input. |
FAQ
What is the Mistral Large 4 model ID?
On Standard Compute it is mistralai/mistral-large-4-0. Mistral's own API uses mistral-large-4. GET https://api.stdcmpt.com/v1/models lists every ID we serve.
Can Mistral Large 4 read images?
Yes. It accepts text and images on Standard Compute, so screenshots from your agent reach it. Video and audio are not supported.
Is Mistral Large 4 open weights?
Mistral lists it as an open-weight model: a mixture of experts with 52B active and 1.05T total parameters. Running it yourself takes multi-GPU server hardware.
Mistral Large 4 or Mistral Large 3?
Large 4 for agent work: Artificial Analysis scores it 38 on its Intelligence Index against 9 for Large 3, and 26.8% against 0% in its Terminal-Bench 4.0 run. Large 3 is cheaper at $0.50 / $1.50 per million tokens and is not in the Standard Compute catalog.
Plans from $19/mo
One key for Mistral Large 4 and the rest of the catalog, with a monthly compute budget you can watch in the dashboard and no overage charges.
Sources: Mistral API pricing · Mistral docs: Mistral Large 4 · Artificial Analysis: Mistral Large 4
Related: Mistral Large 4 benchmarks and price · OpenCode pricing · All model IDs