Mistral Le Chat is now Vibe: The models, plans and prices of Mistral’s AI assistant

Plans, prices and features checked on the official pages on October 25, 2026.

Mistral AI’s chat assistant, long known as Le Chat, now runs under the name Vibe. You still open it at chat.mistral.ai or in the iOS and Android apps, and a Free plan exists, with Pro at $14.99 a month. According to Mistral, existing Le Chat plans, settings and history carry over.

The rename matters because many guides, app listings and search results still say Le Chat. This guide covers what the assistant does today, the models behind it, plans and prices, and the company questions people search for.

Le Chat is now Vibe: What changed and who it is for

On May 28, 2026, Mistral published “Vibe gets to work” and said Le Chat is now Vibe, described as one agent and one license for work and code. Mistral’s help center repeats the line “Le Chat is now Vibe.” The iOS and Android apps are still listed in the stores as “Le Chat by Mistral AI”, according to Mistral’s documentation links.

Vibe is split into two areas in Mistral’s documentation:

• Work is the web and mobile experience. It covers quick conversations and multi-step tasks such as research, drafts, summaries and scheduled runs. A fast/think toggle controls how deeply it reasons.
• Code is for development, in the terminal (Vibe CLI), in a VS Code extension, or as remote sessions on the web at chat.mistral.ai/code. Mistral says these agents can take a task through to a pull request.

The old plain Chat mode is being phased out: Mistral wrote that it will “soon sunset Chat mode”, with history kept in Work mode. Mistral’s documentation also says the rollout is progressive. Free, Pro and Team accounts migrate first, and Enterprise has a six-month window. The older Agents feature has been replaced by Skills, which you call with @.

It suits people who want a general assistant with web search, document work and connectors, developers who want a coding agent under the same subscription, and organizations that care about a European provider with deployment options. Mistral describes itself as an open-weight, sovereignty-minded lab.

What it can do: Search, connectors, documents and voice

From Mistral’s product pages and documentation:

• Web search and deep research: it searches across the web, internal documents and connected tools to compile sourced analysis on focused topics.
• Connectors: more than 100 tools, including email, calendar, Slack, GitHub and Jira, with full MCP compatibility.
• Documents and data: you can upload spreadsheets or connect databases for analysis. The Pro plan adds spreadsheet and slide generation.
• Voice input: powered by Voxtral, Mistral’s speech models. Mistral prices its speech API per minute; our Voxtral Mini Transcribe 3 report covers one of them.
• Approval before sensitive actions: Mistral says Vibe asks first and lets you steer or stop it at any point.
• Temporary chats: not saved to history or used for training, though Mistral notes some data may be kept where legally required.

Mistral’s documentation does not publish usage quotas or rate limits for the plans, so treat any specific message count you read elsewhere as unconfirmed.

Models behind it: Large 4, Medium 3.5 and Small 4

Mistral’s models overview lists three current Mistral flagships:

• Mistral Large 4, an open-weight, general-purpose multimodal model. Mistral says it is a mixture-of-experts design of about 1 trillion total parameters with 52 billion active per token, released October 6, 2026 as a public preview API, with weights promised by the end of the month. For background, see our explainer on mixture of experts.
• Mistral Medium 3.5, a frontier-class multimodal model tuned for agentic and coding use. Mistral’s May 22 post says it powers remote agents in Vibe.
• Mistral Small 4, a hybrid of instruct, reasoning and coding.

The same page also lists third-party Z.ai GLM models. Older models are being retired: Medium 3.1 on August 31, 2026 and Small 3.2 on July 31, 2026. Mistral does not say in the pages read which model answers a given chat, so the model that serves a specific request can change.

Other Mistral models include Mistral OCR (version 4 launched June 23, 2026, version 4.1 appears in the docs), Voxtral for speech, and Devstral for coding. Mistral and Nvidia also announced work on engineering benchmarks, covered in our report.

Plans and prices: Free, Pro, Team and Enterprise

Prices checked on October 25, 2026. The pricing page does not give numeric quotas for any plan.

For developers, Mistral Studio has a free Experiment plan, and API usage is billed per million tokens with separate input and output rates. The pricing FAQ lists Mistral Large at $0.5 per million input and $1.5 per million output tokens, batch processing at 50% off, and cached input tokens cutting input cost by up to 90%. The Large 4 launch post lists $1.36 input and $4.18 output per million tokens. OCR is billed per 1,000 pages and speech models per minute. Compare other providers in our API pricing tracker.

How to get started, log in and get an API key

1. Go to chat.mistral.ai, or install the app from the App Store or Google Play (listed as “Le Chat by Mistral AI”).
2. Create an account or log in. The Free plan needs no payment.
3. Start in Work mode. Type a task, switch the fast/think toggle for harder questions, and add connectors from the connector list if you want email, calendar or GitHub access.
4. For coding, open chat.mistral.ai/code, install the VS Code extension, or use the Vibe CLI in a terminal. Mistral says editors such as JetBrains and Zed work through ACP.
5. For an API key, open Mistral Studio with the same account and use its Experiment plan, then call the models from your own code.

Mistral’s help center at help.mistral.ai covers subscriptions, troubleshooting and data management.

Limits and things to know before relying on it

• Images: the Free plan page lists basic image generation and editing, and Pro lists better generation and editing. None of the models in Mistral’s overview generate images; the page lists OCR 4.1 for reading images and Voxtral TTS for speech. Mistral does not say which model does the image work.
• Transition: features are moving between Chat, Work and Skills. A feature described in an older guide may be renamed or removed.
• Code mode availability: German outlet heise reported on May 29 that Vibe Code was not yet publicly available, while Mistral’s announcement and docs describe it as live. Check your account.
• Limits unpublished: usage caps are not stated, so heavy users should try the Free tier before paying.
• Reputation: searches such as “why is Mistral AI so bad” reflect mixed user opinions, usually about how it compares with the largest US assistants on specific tasks. Judge it on your own workload.

Mistral compared with ChatGPT, Claude and Gemini

Mistral’s selling points are open-weight models, European hosting options and a low-priced Pro plan. The big US assistants generally offer larger feature sets and, in some cases, more mature image and video tools. Check each provider’s pricing page for current numbers, since Mistral’s page lists no usage quotas to compare against.

Mozilla and Mistral also announced in September 2026 that they are bringing open, private and multilingual AI to the browser. Our earlier note on Firefox and chatbot choice shows how that fits.

The company behind it: Founders, funding and ownership

Mistral AI was founded in April 2023 by Arthur Mensch (CEO), Guillaume Lample (chief science officer) and Timothee Lacroix (CTO), according to its about page. It is a French company, and TechCrunch has reported on its data center near Paris. Mistral also announced a Munich hub on September 28, 2026.

On September 8, 2026 Mistral announced a €3 billion Series D at a post-money valuation above €21 billion, to expand compute, research and international growth. Press reports say Samsung led the round, and that ASML led the €1.7 billion round in September 2025 and holds a stake of roughly 11%, a figure attributed to press reports and not confirmed by Mistral in the pages read.

Mistral is privately held, so there is no public stock or stock price. The pages read do not mention an IPO plan or revenue figures. Its revenue sources, according to the pricing and product pages, are subscriptions (Pro, Team, Enterprise), API usage and custom enterprise deployments.

Common questions