Mistral partners with Zhipu AI

Mistral, a French artificial intelligence startup, has begun offering Zhipu AI's GLM model on its platform. This collaboration is part of Mistral's preparations for launching its own large model, Mistral Large 4, which will have 1,000 billion parameters.

Mistral's new revenue strategy

By hosting GLM, Mistral aims to generate recurring revenue by charging companies for each response generated through their services. This approach helps cover the high costs of training AI models, which can be substantial for companies in the field.

Mistral's financial situation

Despite securing significant funding over the years, including 3 billion euros in September 2026, Mistral still faces financial challenges compared to major competitors like OpenAI or Anthropic. This move may help the company strengthen its financial position.

Focus on regional data control

Mistral's decision aligns with the growing emphasis on sovereign inference, where companies prioritize data processing in specific regions. This ensures compliance with local regulations and data privacy laws, a concern for many large French companies.

Testing infrastructure for large models

Hosting GLM allows Mistral to evaluate its infrastructure for handling large models at scale. The company plans to use 4,000 Nvidia Grace Blackwell chips to train Mistral Large 4, a significant technical challenge. GLM will remain in Mistral's offerings even after Mistral Large 4 becomes the default model.