Mistral: a plain-English guide

What Mistral is and who makes it

Mistral AI is a French AI company based in Paris, founded by researchers who previously worked at Google DeepMind and Meta. It builds large language models and an AI assistant, and it is Europe's best-known independent AI lab. Mistral's pitch is control: many of its models are released with open weights, so organizations can run them on their own servers, keep data in their own jurisdiction and adapt the models to their needs.

For everyday users, Mistral's assistant used to be called Le Chat. In 2026 Mistral renamed it Vibe and expanded it from a chatbot into an agent for both office work and coding. Your old Le Chat account, history and settings carried over.

The Mistral model family

Mistral offers general models in several sizes plus specialist models:

• Mistral Large. The flagship. As of October 2026, Mistral Large 4 is in public preview: a very large, natively multimodal mixture-of-experts model. Mistral has said it will publish the weights by the end of October 2026.
• Mistral Medium and Small. Mid-sized and smaller general models that balance quality, speed and cost. Several have been released with open weights.
• Ministral. Small models designed to run on laptops, phones and other devices.
• Specialist models. Models for coding, speech (Voxtral), document reading and reasoning.

Mistral uses a mix of licenses. Many models come under Apache 2.0 or a modified MIT license, which allow commercial use; others are available only through Mistral's own services. The model card for each release tells you which applies.

What Mistral is good at

• European languages. Mistral's models handle French, German, Spanish, Italian and many other languages well, alongside English.
• Self-hosting and customization. Open-weight models can be deployed privately, fine-tuned on company data and run without sending anything outside your infrastructure.
• Data sovereignty. As an EU company, Mistral appeals to businesses and public bodies that want a European supplier under European data protection law.
• Speed. Mistral emphasizes fast responses, and its smaller models are efficient to run.
• Work tasks. Vibe's Work mode handles multi-step jobs across connected tools, such as research, drafting documents and sorting email.
• Coding. Vibe's Code mode works as a coding agent in the terminal, in VS Code or on the web.

Limits to keep in mind

Mistral's models can be wrong or invent details, like all AI models, so check facts that matter; our guide on fact-checking AI answers explains how. Mistral's assistant has fewer consumer extras and third-party add-ons than the largest American apps, and frequent product changes, such as the move from Le Chat to Vibe, can be confusing.

The free plan limits messages, web searches, image generation and coding sessions per day, and the larger models and higher limits require a paid plan. Some newer models, including Mistral Large 4, are still in preview, so details may change. Running the largest open models yourself requires data-center hardware.

How to use Mistral officially

You can use Vibe on the web at chat.mistral.ai and in the mobile apps for iPhone and Android. There is a free plan, Pro for individuals, and Team and Enterprise for organizations. Developers use Mistral's models through its API and developer platform, Mistral Studio, and download open weights from Hugging Face. Large organizations can also arrange private deployments on their own infrastructure, with custom models and agents. If you are new to Mistral, the free plan is enough to test how well it handles your language and your typical tasks before you pay for anything. The official link in the facts section of this page goes to Mistral's website.

Privacy notes

On Vibe's free plan, Mistral may use your inputs and outputs to train its models by default, but you can opt out at any time in your account's privacy settings. On Pro, Team and Enterprise plans, conversations are not used for training by default, and Mistral says data sent through its API is not used for training. If you give feedback with the thumbs up or down buttons, Mistral may use that conversation to improve its models whatever your plan. As an EU company, Mistral is subject to the GDPR, and you can request a copy of your data or delete your account.

If you self-host an open-weight Mistral model, your data stays on your own systems. Our AI privacy and safety guide covers sensible habits for any chatbot.

How Mistral compares

Mistral sits between the closed American assistants and the big Chinese open-weight families. Compared with ChatGPT, Claude or Gemini, it offers more freedom to self-host and a European base, but a smaller consumer app. Compared with Qwen and DeepSeek, it gives European data handling and strong European-language support. Llama remains another open option with a long track record. You can browse all model guides on the models page.

FAQ

What happened to Le Chat?

Mistral renamed Le Chat to Vibe in 2026 and added work and coding agent modes. The web address and your existing account stayed the same.

Is Mistral free?

Vibe has a free plan with daily limits, and many Mistral model weights are free to download. Paid plans and the API cost money.

Is Mistral open source?

Partly. Many Mistral models are open-weight under Apache 2.0 or a modified MIT license, while others are only available through Mistral's services.

Where is Mistral based?

Mistral AI is headquartered in Paris, France, and operates under EU law, including the GDPR.

Is ChatG connected to Mistral?

No. ChatG is an independent AI chat website and is not affiliated with Mistral AI. This page is an informational guide. You can try ChatG free in your browser.