Particle.news
Download on the App Store
4 ARTICLES
·
last mo.

Mistral Launches Europe-Focused Inference Endpoints and 1GW Compute Drive

This gives European firms clearer control over where model inference runs by offering regional API endpoints, a paid Priority tier, and a way to convert multi-year commitments into reserved capacity.

Image
Image
Image

Overview

  • Mistral announced on Tuesday that regional inference endpoints are live with api.eu.mistral.ai for Europe and api.us.mistral.ai for the United States, routing model inputs and outputs to infrastructure in the chosen region.
  • Regional inference carries a 1.1x price multiplier for input tokens, output tokens, and caching, and does not cover control-plane data such as billing, account configuration, or analytics which may be processed outside the selected region.
  • The Priority Tier is now in public preview and offers SLA-backed, pre-committed capacity for mission-critical requests with a published 99.5% uptime target for eligible calls.
  • Mistral will host selected third-party open models starting with Z-ai’s GLM-5.2 under the same regional controls, and it is launching European Compute Units (ECUs) to turn multi-year customer commitments into access to future Mistral-built capacity.
  • The company set a target of building up to 1 gigawatt of European compute by 2030 backed by partner support from Microsoft and NVIDIA plus early corporate sign-ups, but that 1 GW is an aspirational capacity goal and not current operating capacity.