Evanesca vs Mistral Vibe, formerly Le Chat: two French AI chats, two ways of handling your words

At a glance

Each Mistral Vibe cell links to the Mistral page it comes from, read on 6 October 2026; each Evanesca cell links to the page that describes it.

Le Chat is now Vibe

Mistral's help centre puts it simply: “Le Chat is now Vibe, Mistral's unified agent for professional productivity and coding tasks.” Accounts, plans and conversations moved across, and chat.mistral.ai still opens it source. Vibe has three modes: Work, for multi-step tasks across your apps and tools; Code, for developers, in a terminal, an editor or the web; and Chat, the turn-by-turn conversation Le Chat was known for.

This page compares Vibe's plans for individuals — Free and Pro — with Evanesca's chat. Mistral's Team and Enterprise plans change some defaults, noted where they apply.

Same country, different defaults

On jurisdiction, the gap is small. Mistral AI is a French company, registered in Paris source, and its help centre says that, by default, your data is hosted in the European Union source. Evanesca is published by a French company too, and its models run in Europe: in France with OVHcloud or Scaleway, in Mistral AI's own EU region, and in Switzerland with Infomaniak — European companies with no US parent sovereignty.

« Depending on the feature you use, your data can be temporarily transferred outside of the European Union, to the locations listed in the “Subprocessors” tab of our Trust Center. »

Mistral covers those transfers with standard contractual clauses, and Enterprise customers can switch off some of the features involved source. On Evanesca, a message is processed by one operator, in Europe: Evanesca serves no model outside Europe, not even as an option.

Your history, and training

Vibe keeps your conversations on Mistral's side, because that is how it shows your history: zero data retention “is not available for Vibe Work or Chat, regardless of your subscription plan”. Conversations are removed when you delete them or your account, “typically after a short grace period” source.

« Input and output data are used by default to train our artificial intelligence models, unless you opt out of training at any time. »

Files you attach count as input data. You can opt out in your settings, or in the mobile app; a Team administrator can switch training off for the whole organisation, and Enterprise customers are opted out by default source. A thumbs up or down with a comment lets Mistral use that exchange, whatever your plan source.

Evanesca works the other way round. It stores no prompt and no answer — they are held in memory while the answer is written — and your history lives in your browser; of your messages, only billing data, tokens and cost, is kept. Evanesca trains no model. The operators state their own part: OVHcloud keeps “only the data required for billing purposes” source; Scaleway may keep and read requests that cause errors or look abusive, for up to 14 days source; Infomaniak states that queries are “neither recorded nor used to train models” source. And Mistral AI itself, which runs Mistral Large 4, Mistral Small 4 and dictation for Evanesca, keeps API inputs and outputs for 30 rolling days to monitor abuse, and lets its customers object to training — Evanesca's account objects, except for what is sent to Mistral Large 4: it is a preview, and Mistral's commercial terms let it use the data of preview models for training, whatever the objection source source.

Who can read your conversations

Mistral lists three cases in which someone else may see a conversation: content you report to its support team, conversations you share by link, and feedback you send, which Mistral de-identifies before using it to improve its models source.

On Evanesca there is no stored conversation for anyone at Evanesca to open. The operator processes your words to write the answer, and the passport under each answer names it, with what it may keep.

Mistral's models, and the others

Vibe runs Mistral's own models, from the free plan up source. Evanesca serves four of them: Mistral Large 4, in preview, run by Mistral AI itself in its EU region; Mistral Small 4, open-weight, run by Mistral AI itself in its EU region; Mistral Small 3.2, open-weight, run by Scaleway in Paris, with OVHcloud as a fallback; and Pixtral 12B, open-weight, run by Scaleway in Paris. Mistral has not published the weights of Mistral Large 4 yet. Evanesca does not have the rest of Mistral's range.

Next to them, Evanesca's catalogue holds models from Mistral AI, the Swiss AI Initiative, Qwen, DeepSeek, Z.ai, OpenAI, Meta and Google, each showing its checks before you use it, and you can switch model in the middle of a conversation. The comparison tool puts any two or three side by side.

Where Mistral Vibe is the better choice

• Mistral's whole range of models, where Evanesca serves two source.
• More features: over a hundred connectors to your tools, a voice mode and document libraries; its agents work across your apps, while Evanesca's Agent mode works with the web and the conversation's files source.
• A coding agent in the terminal and in VS Code or JetBrains, and apps for iOS and Android source.
• A free plan, a student price, and Team and Enterprise plans with admin controls, SSO and private deployments source.
• Your history follows you from device to device. And it is running today; Evanesca has not opened yet.

If you want one French assistant for everything, from email to code, and you are comfortable with your history being stored, Vibe does more than Evanesca.

Where Evanesca fits better

• No copy of your conversations kept by Evanesca, and no training by Evanesca, with nothing to opt out of privacy.
• An Agent mode that searches the web, writes Word, PDF and Excel documents and creates images, within a spending cap you set, and never runs code Agent mode.
• A choice of models from several makers, run in Europe, each with a public price models.
• No subscription: credit from €2 that never expires, VAT included pricing.
• A passport under every answer: which model wrote it, who ran it, in which country, under which law, and what was kept passport.
• The same models through an OpenAI-compatible API, and for teams a signed Art. 28 DPA business.

Evanesca also reads documents and images, draws charts, handles code and maths, takes dictation and reads answers aloud why Evanesca.

How this page was checked

Mistral's pages were read on 6 October 2026, and its help articles are quoted word for word, in English. Mistral's prices are shown in US dollars, before tax, as its pricing page displays them by default. If a line is out of date, write to us and we will correct it.

• Mistral — Mistral Vibe (formerly Le Chat)
• Mistral Help Center — Le Chat is now Vibe
• Mistral — terms of service for consumers in the EEA, effective 7 August 2026
• Mistral Help Center — where do you store my data?
• Mistral Help Center — do you use my data to train your models?
• Mistral Help Center — opting out of training
• Mistral Help Center — zero data retention
• Mistral Help Center — can other people view my conversations?
• Mistral — pricing

Frequently asked questions