Mistral said its platform will begin hosting third-party open models, starting with Z.ai's GLM-5.2, the company announced on August 12, 2026. The move extends Mistral's open-weights strategy beyond its own models and plants the French lab inside the broader contest for European AI sovereignty.
GLM-5.2 is a third-party open source text model from Z.ai, hosted by Mistral for long-context coding and agentic workflows. Mistral says it is served without modifications. The model carries a one million token context window and a 128,000 token maximum output, with pricing listed at 1.19 euros per million input tokens and 3.74 euros per million output tokens, and 0.119 euros per million for cached input.
The opening is one of three steps Mistral described as it builds what it calls the foundations of customers' AI sovereignty: tighter regional control of inference, broader access to third-party open models on that infrastructure, and a coalition to lock in long-term European compute capacity. Mistral said it plans to build up to one gigawatt of capacity by 2030.
Mistral has positioned open weights as a differentiator for mission-critical work, letting customers inspect, adapt, and keep the intelligence they build. The company said it is contributing to the Open Secure AI Alliance and the Nvidia Nemotron Coalition, and is now extending that openness to models it does not build.
