Mistral Vibe Review (2026) — It Was Called Le Chat Until the 2026 Rebrand, and That Is the Least of the Changes

Freeopen weightsMistral AI

Start with the thing that will confuse you before you get as far as evaluating it: Le Chat has been renamed. During 2026 Mistral rebranded its assistant to Vibe and repositioned it from a chat interface into an agent aimed at work and code. Go to chat.mistral.ai today and the sign-in page reads Vibe, formerly Le Chat.

We have kept this page under the name most people still search for, because search habits outlive rebrands. But if you are hunting for current pricing or limits, they are published under Vibe.

What the rename actually signals

Rebrands are usually noise. This one is a repositioning, and it changes who the product is for.

The old Le Chat competed with ChatGPT on the same terms: a box you type into. Vibe is pitched at doing work rather than discussing it — coding sessions in a CLI, an IDE or the browser, connectors into other systems, scheduled tasks that run without you, a document library it can draw on. Mistral has moved toward the agent end of the market alongside everyone else, and priced accordingly.

If what you wanted was a European alternative to ChatGPT for everyday questions, that product still exists inside this one, but it is no longer what the marketing is aimed at, and the free plan is now shaped like an evaluation allowance for the agent features rather than a generous daily chat quota.

What you get without paying

Mistral's pricing page lists the free plan as: limited messages and web searches, limited coding sessions, image generation, access to Mistral Studio for testing models, a monthly allowance of API credits, over a hundred connectors, up to five scheduled tasks, and a limited document library.

The word doing the heavy lifting there is limited, which appears four times and is never quantified. The API credit allowance is the one concrete number, and it is the most interesting part of the free plan for anyone technical — a genuine monthly budget to build against without a card, which is more than most competitors hand out.

Above free, Pro runs at 14.99 US dollars a month, with a verified-student rate of 5.99, and Mistral describes it as up to forty times the free allowance on messages, searches and images. A forty-fold multiplier is a useful signal about how small the free tier is, even without the base number.

The open-weights claim, read carefully

Mistral's reputation rests on publishing open-weight models, and that reputation is deserved — but the licensing has a condition that a lot of coverage skips.

Mistral's own pricing page states that its open-weight models are Apache 2.0 licensed for research and individual use, while commercial deployments require a Mistral licence. Apache 2.0 does not normally work that way, so what is being described is a split arrangement rather than a plain permissive release, and not every model behind the hosted product is published at all.

For an individual, a researcher or a hobbyist, this is a real freedom and one of the better offers in the market. For a company planning to self-host and ship something, it is a licensing conversation, not a download. Worth knowing before you build a plan on the word open.

Pros

• European vendor and European data-protection regime, which for some organisations decides the question on its own.
• A real monthly API credit allowance on the free plan, not just chat messages.
• Open-weight models that individuals and researchers can genuinely run themselves.
• Agent features, connectors and scheduled tasks available to try before paying.
• Student pricing at well under half the standard rate.

Cons

• The rename to Vibe fragments the documentation and everything written about it before 2026.
• Free-plan limits are described only as limited, with no published figures.
• The Apache 2.0 claim carries a commercial-use condition that is easy to miss.
• The shift toward agents makes it a heavier tool than someone wanting a simple assistant needs.
• Smaller ecosystem and less third-party integration than the American incumbents.

Best alternatives

• HuggingChat — a more thoroughly open option, self-hostable end to end rather than at the weights layer only (chatbots).
• Claude — stronger on long documents and nuanced reasoning if jurisdiction is not your constraint (chatbots).
• DeepSeek Chat — MIT-licensed weights with no commercial-use carve-out, though hosted in a jurisdiction with its own implications (chatbots).

FAQ

Is Le Chat the same thing as Mistral Vibe? Yes. Mistral renamed Le Chat to Vibe during 2026 and repositioned it from a chat assistant to an agent for work and code. The URL chat.mistral.ai still takes you there and the sign-in page now reads Vibe, formerly Le Chat. If you are searching for Le Chat pricing or Le Chat limits, the current information is published under the Vibe name.

What does the free plan include? Mistral's pricing page describes the free plan as limited messages and web searches, limited coding sessions, image generation, access to Mistral Studio for testing models, a monthly allowance of API credits, over a hundred connectors, up to five scheduled tasks and a limited document library. The numbers behind the word limited are not published, so the free tier is best treated as an evaluation allowance rather than a dependable daily quota.

Are Mistral's models really open source? Partly, and the caveat matters. Mistral publishes open-weight models under Apache 2.0 for research and individual use, but its own pricing page states that commercial deployments require a Mistral licence. Not every model powering the hosted product is released this way either. Open weights here means genuinely useful freedom for individuals and researchers, not a blanket permission for a business to self-host and ship.

← Back to all free AI tools · Visit Mistral Vibe ↗