Mistral AI

Mistral AI is a Paris-based frontier AI company offering both a developer API platform and Vibe, an AI agent for consumers and enterprises that handles chat, work automation, and coding in one product. Vibe (rebranded from Le Chat in May 2026) runs on Mistral's own model family, from lightweight Ministral models to the flagship Mistral Large 3 and reasoning-focused Magistral line, and is available via web, iOS, Android, a terminal CLI, and a VS Code/JetBrains extension. Its two newest capabilities, Work Mode and Code Mode, mark a shift from a conversational assistant into an execution agent: Work Mode reads emails and calendars, drafts documents in a Canvas editor, runs scheduled workflows, and pushes outputs to tools like Notion, SharePoint, or Slack, always asking for explicit approval before any action that modifies data, while Code Mode runs remote coding sessions in isolated sandboxes and delivers finished work through to a pull request.

For developers, Mistral Studio provides a platform to build, test, and deploy custom AI agents and applications with agent orchestration, end-to-end observability, evals, guardrails, and full deployment portability across self-hosted, EU-hosted Mistral cloud, or major cloud provider infrastructure. The API catalog spans more than 30 models across text, reasoning, coding (Devstral, Codestral), voice (Voxtral), OCR, embeddings, and classification, priced per million tokens with batch processing available at a 50% discount. Mistral Forge adds tooling for training, aligning, and evaluating fully custom models on proprietary data, while Compute offers the underlying GPU infrastructure Mistral uses to train its own models as a service to enterprise customers.

A distinctive part of Mistral's strategy is open-weighting even its flagship models, including Mistral Large 3, Medium 3.5, and Small 4, under licenses that allow free research and individual use while requiring a commercial license for production deployment. Combined with EU-based hosting and self-hosted deployment options, this makes Mistral a common choice for European public-sector, financial services, and defense customers with strict data-sovereignty requirements, alongside named enterprise customers like ASML, HSBC, CMA CGM, Stellantis, and the European Patent Office.

Founded in April 2023 by Arthur Mensch (CEO, ex-Google DeepMind), Guillaume Lample (Chief Scientist, ex-Meta), and Timothée Lacroix (CTO, ex-Meta), Mistral AI has raised over $5.5 billion in total funding, including an $830 million debt round in March 2026 to fund a Nvidia-powered data center outside Paris, and reported a valuation near $23 billion (roughly €20 billion) in 2026. The company employs around 350 people and reported approximately $400 million in annual recurring revenue in 2026, positioning it as Europe's most valuable independent AI lab.

Vibe Free offers limited access to Mistral's SOTA models, web and mobile access, limited messages, web searches, and coding sessions, image generation, and 100+ connectors. Vibe Pro is $14.99/month with more messages, web searches, more complex task handling, all-day coding in the CLI, IDE, and web, more image generations, and chat and email support. Team is $24.99/user/month, adding up to 30GB of storage per user, domain name verification, and data export. Enterprise offers custom models, custom agents, custom workflows, audit logs, SAML SSO, and white-labeling via a private deployment, priced on request. A Student plan offers Vibe Pro for $5.99/month for verified students. Separately, the API is billed per million tokens: for example Mistral Large 3 costs $0.5 input / $1.5 output, Medium 3.5 costs $1.5 input / $7.5 output, and Small 4 costs $0.15 input / $0.6 output, with batch processing available at a 50% discount and Enterprise APIs available at a 75% premium for regional controls, SLAs, and premium support.