Mistral Enterprise: Features

Updated November 2, 2026 — Le Chat is now Vibe. As of May 28, 2026, Le Chat has been rebranded as Vibe (still at chat.mistral.ai): a single agent for work and code that preserves your conversations, settings, and subscriptions. The legacy Le Chat experience remains available in Chat mode. In Work mode, agents are now skills (invoked with @), memories serve as the knowledge base, and deep search is the Deep Research skill. Plans are now named Mistral Free, Pro, Team, and Enterprise. This lesson uses the name Vibe; note that some menus described here may have changed names or locations.

A Distinct Product, Not an “Enhanced” Version

Mistral Enterprise isn’t just the consumer Chat with extra options toggled on. It’s a product architected from the ground up to meet organizational constraints. Data security, access control, auditability, and integration with existing infrastructure aren’t afterthoughts; they structure the product. You see the difference on the very first screen: while a consumer user creates an account with an email address, an employee arrives already authenticated by their company’s identity system.

Authentication and Identity Management

Mistral Enterprise integrates natively with your identity providers via SAML 2.0 and OIDC. Your employees log in with their usual credentials—Azure AD, Okta, Google Workspace, Ping Identity—without creating a new account. The practical implications matter more than the protocol: the day an employee leaves the company, deactivating their account in the directory immediately revokes their Chat access. No administrator needs to remember to disable this specific tool. Your existing password policies and multi-factor authentication rules apply as-is. Provisioning and deprovisioning happen automatically via SCIM: a new hire added to the Marketing group in the directory gains access on their first morning.

Access relies on three roles, which you should assign before opening the service. The Administrator holds global configuration, user management, data policy, and audit rights. This is the role organizations often forget to designate, leaving it by default to whoever opened the tenant. The Team Manager sees usage within their scope and manages allowed models, letting them reserve expensive models for profiles that actually need them. The User accesses Chat within the limits set by the administrator.

Data Policy and Zero Retention

Under zero retention, Mistral AI stores no conversation data beyond the session. Prompts and responses aren’t stored, aren’t used to train models, and aren’t accessible to Mistral teams. Committees often misunderstand this point while hunting for a checkbox to tick: on the Enterprise plan, this behavior applies by default. The legal counsel pasting a contract draft into Chat doesn’t need to wonder if that text survives somewhere on a server.

Administrators can adjust this behavior if needed. Limited retention keeps exchanges for 7, 30, or 90 days when audits or usage tracking require revisiting a conversation. Retention with anonymization offers a compromise: it keeps usage metadata while deleting content, allowing you to steer adoption without retaining confidential material.

The Administration Console

The console shows real adoption metrics: daily, weekly, and monthly active users; query volume by team and model; average response time; satisfaction rates; and trends over time. If usage stagnates three months after launch, that signals a training gap, not a lack of interest. Administrators also control accessible models here: they enable or disable Mistral Large 3, Medium 3.5, Codestral, and others; set default models by team; and restrict access for cost or internal policy reasons.

Every action is logged in an exportable audit trail covering logins and logouts, configuration changes, connector and agent usage, and administrative actions like adding or removing users and changing roles. These logs make compliance demonstrable rather than just declarative.

Productivity Features

Agents are preconfigured assistants with a system prompt, context, and instructions tailored to a specific business use case. A legal agent analyzes contracts against your standard clauses and flags deviations. An HR agent answers frequent questions about internal policy, relieving a team that otherwise answers the same leave question ten times a day. A sales agent generates proposals from your product catalog. Shared at the organization level and managed by administrators, agents are built once, not by every employee.

Canvas mode addresses the need for long documents: drafting, editing, rephrasing, and translation happen directly in the interface. This eliminates the copy-paste back-and-forth that breaks formatting and creates version errors. Additional features include uploading PDF, Word, Excel, or image files for analysis; web search to enrich responses with current data; and integration with corporate tools via Enterprise Search, which Lessons 5 through 8 detail.

What Sets Mistral Enterprise Apart

Compared to ChatGPT Enterprise, the comparison hinges on three gaps: European hosting versus American, zero retention by default versus optional, and native GDPR compliance versus retrofitted adaptation. These are exactly the points a DPO will grill you on. Against Copilot, the argument is independence: Vibe works with any identity provider without assuming a Microsoft ecosystem, the models are open and transparent, and you avoid de facto lock-in to the Office suite. Compared to an on-premise deployment, the trade-off is cost and effort: no GPU infrastructure to fund, no model updates to manage, and support plus SLA included. The decision ultimately rests on your actual sovereignty requirements, the subject of the next lesson.

Key Takeaways

• Mistral Enterprise offers SSO/SAML, zero retention by default, and a comprehensive administration console
• Audit logs provide the traceability required by regulations
• Custom agents transform Chat into a specialized business tool
• European hosting and native GDPR compliance are major differentiators against US competitors