Customers can direct requests to a European or a US endpoint. The commitment covers model execution, not agents, batch processing, file management, billing or API keys.
Mistral has made regional inference generally available. Via api.eu.mistral.ai, model execution is routed to data centres in EU and EFTA states; via api.us.mistral.ai to the United States. The standard endpoint api.mistral.ai is explicitly not region-bound and carries no commitment about where processing occurs. The premium is ten percent.
The distinction between storage location and processing location is the core of the news. Mistral says it already hosted customer data in the EU. What is new is a verifiable commitment about where computation happens. The company itself sets out the limits: of the additional tools, only function calling runs regionally; agents, batch processing and file management are absent from the regional endpoints. Account settings, API keys, billing and usage statistics may be processed outside the chosen region.
In parallel Mistral is introducing a Priority Tier with precedence during congestion and a 99.5 percent uptime commitment, at 1.75 times standard pricing and only via contract. The vendor is also admitting third-party open models, starting with GLM-5.2 from Chinese provider Z.ai.
What this means for decision-makers
- Distinguish explicitly between storage location and processing location in your contracts – only the latter is newly committed here.
- Check whether your application needs agents, batch processing or file storage; the regional commitment does not cover them.
- Budget the ten percent regional premium, and where applicable 75 percent for priority, into your running costs.
This story was produced automatically from the source named above and checked by software before publication. The image is symbolic and shows neither the event nor a real person. How this paper is made