We target the performance-cost frontier: our models always outperform models of similar size.Mistral large, our latest flagship model, achieves top-tier performance on independent benchmarks and is served at high speed.Check out our pricing for more details.
Use our generative models where your business needs them. Whether on our public platform, through our preferred cloud partners, or in your own environment (private cloud or on-prem), we provide a compliant access to our technology.
Customize Mistral AI models for your business. We offer different ways to easily and efficiently adapt Mistral AI models to your specific needs, reducing the cost and expertise required for customization at scale.
We do not use your data for training on our public platform. With our self-deployment and cloud channels offerings, we don’t even see your input data.
Your data is encrypted at rest (AES256) and in transit (TLS 1.2+).
Our models are easy to control to detect certain kind of user input and ban certain kind of output. Le Chat illustrates how to make moderated applications with Mistral models.
Our models are multilingual by design; benchmarks show that our models are less biased than the competition. We help your business make editorial choices regarding model outputs – we don’t force ours.
Discover the roster of companies already using our services
Use our Mistral models to process, summarize, classify or translate any kind of text. Summarize a long report, classify customer reviews, translate emails into other languages, generate marketing campaigns - the possibilities are endless!
Give your employees easy access to internal company knowledge by building retrieval-augmented (RAG) applications. You can use our state-of-the-art embedding model for that, as well as the function calling capacities of our models.
Leverage your developers with a custom-built coding co-pilot and accelerate your coding speed! Mistral models have proven to be particularly strong in coding, making them a great asset to accelerate application development or even IT legacy modernization.
Our multilingual models can be tuned to follow a certain editorial line and used to generate content as you see fir for your customers.
