Mistral Large 4
Large language model from Paris-based Mistral AI for text and images with a context of one million tokens, available via API and as open weights.
What it does
Mistral Large 4 is Mistral AI’s new flagship model: a mixture-of-experts model with around one trillion parameters that handles text and images and can take up to one million tokens into account at once. It can answer directly or reason step by step, and it understands more than 160 languages, including every official EU language. The API offers function calling, structured outputs, document Q&A, batch processing and agents. Mistral names coding, cybersecurity, finance and law as its strengths.
Who it suits
Development teams and companies that want to build a powerful European model into their own applications, agents or analyses – for example long contracts, code reviews or documents with charts.
Data protection for businesses
Data is stored in the EU by default. Depending on the feature, it may be transferred temporarily to sub-processors outside the EU. A dedicated EU endpoint keeps processing in the EU, at 10% extra cost. Mistral provides a data processing agreement. Even on paid API access, inputs may be used for training unless you opt out under “Privacy” in the admin console. Enterprise contracts are available for companies.
Limits
Since 5 November 2026 the model has been generally available via the API. The weights are on Hugging Face under the “Mistral Ecosystem Open-Weight” licence, a modified MIT licence; according to Mistral, anyone offering the model as a service to others needs a separate commercial licence. Running it yourself requires very substantial computing power.