Mistral Integrates China's GLM-5.2 Model into the 'European Sovereign AI' Platform

According to Beating Monitor, Mistral has long championed “European sovereign AI,” yet it has now directly integrated a Chinese model into its platform. Mistral has announced it will begin hosting third-party open models, with the first being GLM-5.2 from Z.ai, a subsidiary of Zhipu AI. The model runs directly on Mistral’s infrastructure, enjoying the same regional controls and service guarantees as Mistral’s own models. Mistral has previously been the subject of jokes within the community, comparing it to DeepSeek. After the release of Mistral Large 3, researchers analyzed its architecture and found it closely mirrors DeepSeek V3/V3.1, with only minor adjustments such as expert configuration. Now, Mistral has openly admitted integrating a Chinese model into its “sovereign AI” framework. Some in the community have joked that European sovereign AI teams no longer need to repackage Chinese models—just host them on European servers. Jokes aside, Mistral has redefined “sovereign AI”: the model doesn’t need to be trained by a European company; what matters is that enterprises can choose their own models and control where they run, where data stays, and who holds the compute power. Thus, a Chinese model running on European infrastructure can still qualify as “European sovereign AI” under Mistral’s definition.