Le Chat

At a glance

Le Chat is now Vibe, Mistral AI’s hosted assistant for conversation, research and multi-step work. The existing chat.mistral.ai entry point remains. It is worth trying on a familiar, low-stakes task if you want a web assistant that can move from an answer to a structured deliverable. Judge it by the result you can verify and reuse, rather than the promise of an autonomous colleague. This guide covers the web experience; coding through the CLI or editor is a separate mode.

Good fit

People who want drafts, sourced research or summaries in one conversation and are willing to review the output. Current documentation merges quick chat and deeper Work into a fast/think experience, while Code handles development tasks. Connected email, files and other services can supply context, but are optional for an initial public-information trial. Their usefulness depends on which tools and permissions your workspace makes available, not just which integrations appear on a marketing page.

Consider something else when

You require offline processing, guaranteed factual answers or a paid individual plan that automatically excludes every interaction from training. The documented individual Vibe service is not opted out of model training by default; attached documents count as inputs too. Enterprise has different defaults, and organization controls vary. A hosted assistant is also a different choice from downloading model weights and operating them yourself. Connecting private services should follow a clear need for that access.

Start here

Open chat.mistral.ai and review the terms and account steps presented. The inspected session showed Vibe Chat, Fast and a terms-acceptance dialog; it stopped before acceptance or a model response. For a proposed first trial, ask for a short comparison of two public product plans, with a source for each factual row and unknowns marked explicitly. Check the sources before reusing the table. Before uploading private work, review the training control under Admin, Manage, Vibe, Privacy. The separate Studio/API opt-out does not change the Vibe setting.

Cost and requirements

The current pricing page offers a limited Free plan. On 31 October 2026 its US-dollar list prices show Pro at $14.99/month and Team at $24.99/user/month, with tax treatment and allowances depending on the selected offer. Paid plans raise usage limits; they do not promise unlimited access. Some plans include API credits, while extra pay-as-you-go usage can add cost. Check the actual allocation and billing setting instead of treating a web subscription as unlimited API usage. The web app needs no native installation; mobile and Code have their own setup.

Limits to know

This assessment inspected the public entry screen and read current product, pricing, training and export guidance. It did not assess answer quality, research completeness, connector actions or coding. The documentation describes a progressive migration, so accounts may show different navigation. Temporary Chat is documented as excluded from history and model training, with legally required retention exceptions; that is not a universal zero-retention guarantee. A personal-data export is available, but its completeness and portability were not tested.

Related work

• Mistral AI: provider, models and related research (context)