Mistral AI restructured its subscription tiers on October 12, 2026. The Team plan no longer includes the Admin API, as this feature now resides exclusively in the Enterprise tier. The Pro plan monthly API credit limit dropped 15% from $30 to $25.50, although the tier added spreadsheet and slide generation plus faster responses. The Free tier now includes Vibe CLI and Vibe for IDE, which were previously limited to Pro and above. For the Timestral forecasting models, which include 5 million, 28 million, 123 million, and 209 million parameter versions, a Modified MIT license applies. This license excludes any company or employer with global consolidated monthly revenue exceeding $20 million from using the weights under those specific terms, requiring those users to request a commercial license from Mistral instead. The Timestral models can forecast related series together and accept extra information called covariates. These models generate predictions in 21 quantiles, providing predicted levels at different percentiles rather than a single answer. The Vibe interface provides mobile support for iOS and Android. It includes tools for web access, coding in the CLI, and a VS Code or JetBrains plugin. Users can create and load instructions, tools, and scripts for repeatable tasks. The Vibe models can also handle voice mode for recording, transcribing, and summarizing audio. Arthur Mensch, the Mistral AI CEO, claims that more than 50% of current enterprise software could be replaced by AI tools through a process of replatforming.
Data sovereignty and Microsoft infrastructure
Microsoft is spending billions of dollars on Mistral’s computing infrastructure in Europe. Through this agreement, Microsoft Azure customers can use Mistral’s data centers in France to develop software. Mistral provides Medium 3.5 and OCR 4 models to Microsoft Foundry. Because Mistral is headquartered in Paris, it does not face the US CLOUD Act. US authorities cannot compel Mistral to hand over data stored on its servers, unlike with US-based companies like OpenAI or Google, which are subject to the US CLOUD Act. All Mistral services run on EU-based, ISO 27001-certified data centers. This partnership aims to deliver sovereignty while allowing access to US software and security features. Businesses with independent data centers that access Microsoft services via Azure Local can run Mistral’s open models. This deal underscores the growing interest in reducing dependence on US technology so other countries can maintain greater control over their future economy. You should prioritize this if your business requires data to stay within European jurisdiction.
| Feature | Mistral Small 4 | Gemini 3.6 Flash |
|---|---|---|
| Input (per 1M tokens) | $0.15 | $1.50 |
| Output (per 1M tokens) | $0.60 | $7.50 |
Pricing efficiency and the move toward control
Mistral Small 4 costs $0.15 per million input tokens and $0.60 per million output tokens, whereas Gemini 3.6 Flash costs $1.50 per million input tokens and $7.50 per million output tokens. Mistral Medium 3.5 models cost $1.50 per million input tokens and $7.50 per million output tokens. In 2026, enterprise AI adoption relies on governance and reliability rather than simple speed. Gartner reported that nearly 40% of enterprise apps will feature task-specific AI agents in 2026, which is a large increase from the less than 5% seen in 2025. Senior leaders indicate that 79% of their companies have started using AI agents for their operations. CFOs have dedicated 25% of their total AI budget to AI agents. The industry is moving from experimentation to selection in 2026. This shift favors disciplined operators who demand FinOps integration and performance benchmarks before approving scale. As AI systems influence decisions that affect customers, pricing, or eligibility, leaders become aware of personal and institutional accountability. The removal of the Admin API from Team plans restricts the ability of mid-sized firms to manage users programmatically without upgrading to Enterprise. Does this revenue-based restriction for Timestral prevent smaller European developers from accessing advanced forecasting?
