Microsoft has made a notable shift in its AI strategy, favoring Anthropic’s Claude 4 over OpenAI’s GPT-5 for its Visual Studio Code auto model feature. This decision is based on internal benchmarks that highlighted the superior performance of Claude Sonnet 4 in GitHub Copilot, according to an internal email from Julia Liuson, head of Microsoft’s developer division. The move suggests a quiet but clear preference for Anthropic’s models.
In addition to this, Microsoft is making significant investments in its own AI research and development. Mustafa Suleyman, Microsoft’s AI chief, mentioned that while the MAI-1-preview model was only trained on a small cluster of 15,000 H100s, the company is expanding its own AI infrastructure. This indicates a strategic effort to develop proprietary AI models alongside partnerships with external providers.
Furthermore, Microsoft is reportedly planning to integrate Anthropic’s AI models into some features of its Microsoft 365 apps. The Information reports that the Microsoft 365 Copilot will be partly powered by Anthropic models, following findings that these models outperformed OpenAI’s in specific applications like Excel and PowerPoint. This shift may signal a broader trend in the industry where companies are diversifying their AI provider strategies.