Mistral Large 4

Mistral Large 4, also called ML4 in Mistral's announcement, is a public-preview general-purpose multimodal Mixture-of-Experts model. Mistral says the preview API is available in Mistral Studio from October 6, 2026; the model weights are planned for release by the end of the month, so they are not yet available for self-hosting.

What it offers

Large MoE design: Mistral documents 1.05 trillion total parameters, 49 billion active parameters, and a 1.6 billion-parameter vision encoder.
Multimodal and long context: the API documentation lists image-capable input and a 1M-token context window.
Agent-facing APIs: the preview supports chat completions, conversations, function calling, structured outputs, document Q&A, batching, and Mistral's Agents and Conversations APIs.
Open-weight roadmap: Mistral calls the model open-weight, but its announced weight release is future tense. Treat the current product as an API preview until that release arrives.

Who should evaluate it

Teams building coding agents, document workflows, or image-aware assistants can test ML4 through the preview API, especially if a long context window matters. Security and regulated-workload teams should assess it under their own controls once weights and deployment guidance are actually published. Its public-preview status makes it unsuitable for assuming production stability without a fallback model.

Pricing and availability

Mistral's model page lists US pricing of $1.36 per million input tokens, $0.14 per million cached input tokens, and $4.18 per million output tokens. It also shows a lower “Batch” column: $0.68, $0.07, and $2.09 respectively. Verify the live pricing page before procurement because preview pricing and availability can change.

Quick start

Open the official Mistral Large 4 model page.
Try mistral-large-4 in Mistral Studio or through the documented API endpoints.
Test a bounded coding, document, or visual task against your current model and record cost, latency, and failure cases.
Keep a fallback in place while the preview and planned weight release evolve.

Alternatives

Mistral Small is the older Mistral Small v24.09 catalog entry.
Claude Opus 5.5 is a hosted frontier-model alternative.
GPT-6 Astra is another hosted model for agent evaluation.

FAQ

Are the weights available now?

No. Mistral announced a public API preview and said it plans to release weights by the end of October 2026.

Does it accept images?

Yes. Mistral documents multimodal input and a vision encoder; validate the supported formats and limits in the live API reference.

Is it ready for production?

It is a public preview. Evaluate it with monitoring, limits, and a fallback before treating it as a stable production dependency.

Conclusion

Mistral Large 4 is a notable new Mistral preview for teams comparing multimodal, long-context agent models. Its API is usable now, but the promised open weights and the final deployment picture remain pending.