Rebrand and enterprise functionality
Mistral AI renamed its Le Chat assistant to Vibe on May 28, 2026, and expanded its functionality to include Vibe Work, Vibe Code, Vibe Chat, and scheduled tasks. I find the Enterprise tier the only logical choice for organizations that require on-premises deployment or a private cloud to keep data within their own environment. While the Team plan costs $24.99 per user per month with a $50 monthly minimum, Enterprise requires a custom quote from Mistral sales. Enterprise customers get SAML SSO, audit logs, white label capabilities, and custom models. Unlike individual plans where training is on by default, Enterprise customers are opted out of model training automatically. You already know that companies need strict data controls. Vibe Work plans multi-step tasks across tools like Google Workspace, Outlook, SharePoint, Slack, and GitHub and shows the plan for approval before execution. Vibe Code provides coding agents in isolated sandboxes and includes plugins for VS Code and JetBrains. The Pro plan costs $14.99 and gives users 15GB of storage and 40x image generation, while the Team plan gives 30GB of storage per user and domain name verification. The Free plan provides 100+ connectors, 5 scheduled tasks, and $10 in monthly API credits. Verified students get the Pro plan for $5.99 per month for up to twelve months and receive $15 in monthly API credits.
Pricing rivalry and data risks
The pricing landscape changed this month because OpenAI’s GPT-6 Luna undercuts Mistral Large 3 on both input and output costs. Mistral Large 3 costs $0.50 per million input tokens and $1.50 per million output tokens, yet GPT-6 Luna costs only $0.10 and $0.50 respectively. Every Vibe plan now bundles monthly API credits: $10 for Free, $30 for Pro, and $50 for Team.
| Feature | Free | Pro | Team | Enterprise |
|---|---|---|---|---|
| Price | $0 | $14.99 | $24.99/user | Custom |
| API Credits | $10 | $30 | $50 | Custom |
| Training | Opt-in | Opt-out | Opt-out | Opt-out |
| Deployment | Web/Mobile | Web/Mobile | Shared | On-premises/Private |
I find the Microsoft/Azure partnership problematic because routing EU data through US-owned hyperscalers exposes it to US extraterritorial law like the CLOUD Act, which allows US authorities to access data regardless of the original location of the servers. Small businesses often prefer the Team plan because it gives 200 Flash answers per day compared to the 150 provided in the Pro plan. DeepSeek V4.1 Flash remains a competitor with off-peak pricing of $0.15 per million input tokens and $0.60 per million output tokens. The Samsung investment in September 2026 at a post-money valuation of over €21 billion reduces the risk of the company disappearing during a contract.
Model performance and deployment
Mistral Enterprise runs on Mistral Medium 3, which differs from the flagship Mistral Large 3 released on December 2, 2025. Mistral Large 3 uses a 675-billion-parameter sparse Mixture-of-Experts architecture with 41 billion active parameters per token and provides a 256,000-token context window for complex reasoning tasks. Large 3 achieves 85.5% on MMLU, 92% on HumanEval, 43.9% on GPQA Diamond, and 23.8% on SimpleQA. Mistral Small 4, the March 16, 2026 release, unifies reasoning, vision, coding, and text processing into a single 119-billion-parameter model. This model has 128 experts and 4 active per token, and it gives a configurable reasoning_effort setting that ranges from low-latency replies to deep chain-of-thought. Vibe Code uses Devstral 2 to pull context from GitHub, GitLab, Jira, and Linear. Does the complexity of managing custom models outweigh the benefits of the specific Vibe Work workflows? I recommend the Enterprise plan for organizations requiring total data sovereignty.
