Mistral Offers Guaranteed EU Data Processing and Priority Tier with Important Restrictions

Mistral has announced regional inference, generally available now, allowing customers to send requests to EU (api.eu.mistral.ai) or US (api.us.mistral.ai) endpoints, with a 10% surcharge. However, only function calling runs regionally among add-on tools; agents, batch processing, and file management are missing. Also, the model offering may differ by region. Account settings, access keys, billing, and usage statistics may be processed outside the chosen region, and limited secure transfers to subprocessors outside the region are mentioned. For customers needing text processing, the EU endpoint works, but for agents or Files API, the same promise is not given. Mistral also introduced Priority Tier, currently in open beta, which provides priority processing during congestion for a 75% surcharge, plus a 99.5% uptime SLA. The priority can be activated via a service_tier parameter, and customers get individually negotiated rate limits. Discounts from prompt caching apply before the surcharge. Additionally, Mistral is opening its platform to open models from other providers, starting with GLM-5.2 from Chinese vendor Z.ai, and is collecting long-term commitments from large customers as European Compute Units to justify building new data centers in Europe.