Mistral AI is opening its inference platform to third-party open models, beginning with GLM-5.2 from Chinese developer Z.ai. The French company says customers will be able to use external models on the same infrastructure, regional settings and service levels offered for Mistral’s own models.
The move is part of a broader infrastructure push. Mistral has made regional inference endpoints generally available, allowing customers to choose whether AI processing takes place in Europe or the United States. It has also introduced a Priority Tier in public preview, with tailored rate limits and an uptime service-level agreement for critical business applications.
According to Mistral, these features help organizations meet data residency, compliance and reliability requirements. However, the company notes that limited transfers to sub-processors outside the chosen region can still occur. VentureBeat reports that this may affect functions such as web search and external tool calls, which customers could restrict or disable.
Long-term capacity commitments
Mistral is also asking European companies to commit to compute capacity over several years through European Compute Units, or ECUs. These commitments are intended to support the construction of 200 megawatts of European infrastructure by the end of 2027, with a target of up to one gigawatt by 2030.
ECUs can be used for inference, model adaptation, training and managed computing services. VentureBeat reports that Mistral seeks commitments of around five years and describes the contracts as non-cancellable.
The company is positioning itself less as a model provider and more as a regional AI infrastructure operator. Its approach combines access to open models with locally controlled deployment, while relying on major enterprise customers to finance new data-center capacity.
Sources
- Mistral says its platform will support third-party open models, starting with Z.ai’s GLM-5.2, and run them on the same infrastructure as its own models – Mistral AI Blog
- Mistral AI wants to build 1 gigawatt of European compute by 2030 — and lock in customers now – VentureBeat
Stay up to date
AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:
