Mistral logo
Models

Mistral

European language models, open-weight and hosted. Mistral AI is a French company that publishes language models, some with openly downloadable weights and others available only through its hosted API. The range covers small efficient models, mixture-of-experts models and models tuned for code. Teams use them for chat, extraction, code assistance and retrieval-augmented generation.

When it is the right choice

  • You want a European model provider for data residency or procurement reasons
  • You need efficient models that run well on modest GPU hardware
  • You want the option to start on the hosted API and move to self-hosting later
  • Not the right fit if your workload depends on a feature only another provider offers; compare on your own prompts first

How we use it

  • Compare Mistral models against alternatives on your real tasks before committing
  • Self-host open-weight Mistral models with vLLM on Kubernetes, or wire up the hosted API behind a gateway
  • Track cost, latency and output quality per model so switching is a measured decision
  • Check the licence of each Mistral model, as terms differ between models
Works well with

Often used together

Questions

Mistral: frequently asked

Are all Mistral models open-weight?

No. Some are released with downloadable weights and others are available only through Mistral's API. Each model has its own licence, so check it for the model you choose.

Can we run Mistral inside the EU?

Yes. Open-weight models can run in any EU cloud region you choose, and Mistral's own hosted service is operated from Europe. We help you confirm where data is processed for your compliance needs.

Running Mistral, or deciding whether to? Tell us what you have; the first consultation is free.

Let's talk »