In-Region Inference, Open Models, and New European Infrastructure for Sovereign AI

Mistral is strengthening sovereign AI infrastructure to allow enterprises and nations to directly control AI models, execution locations, and computing capacity. The core pillars are enhanced in-region inference and stability, support for third-party open models, and securing AI computing capacity in Europe.

With Mistral Regional Endpoints, now generally available, users can select Europe or the United States as the inference execution region. Inference and related processing occur in the selected region to meet data residency, regulatory, and latency requirements, though limited and protected out-of-region transfers to some sub-processors may occur.

Mistral Priority Tier for mission-critical workloads is also available in public preview. It offers customized request limits and committed service levels, ensuring stable capacity even during peak demand based on uptime SLAs.

Mistral is expanding its platform beyond its own models to include third-party open models. The first supported model is GLM-5.2 by Z.ai, and future open models will be offered under the same infrastructure, regional controls, and service agreements as Mistral models to broaden model choice.

Mistral emphasized that open-weight models allow users to inspect model internals, fine-tune them for specific purposes, and directly own the intelligence they build. It is also participating in the Open Secure AI Alliance and the NVIDIA Nemotron Coalition to expand the open frontier model ecosystem.

The company also presented a plan to unite European enterprises and institutions to secure long-term AI computing capacity, aiming to build up to 1 GW of capacity by 2030.