At the Build developer conference, Microsoft unveiled seven new AI models developed entirely in-house, marking a significant step towards reducing dependence on external partners like OpenAI and Anthropic. This initiative supports Microsoft's strategy to achieve long-term self-sufficiency in AI technology.
Among these models, the standout is MAI-Thinking-1, a reasoning system claimed to perform comparably to Anthropic's Claude Sonnet 4.6 in blind tests, and it meets the standards of the more advanced Claude Opus 4.6 on coding benchmarks. CEO Mustafa Suleyman emphasized that this model was developed without using distillation techniques from competitors, appealing to enterprises focused on data integrity.
Additionally, Microsoft introduced MAI-Code-1-Flash, a coding model with 5 billion parameters, which is now being integrated into Visual Studio Code and GitHub Copilot. Another new model, MAI-Image-2.5, is recognized as the second best in its category, further enhancing Microsoft's AI offerings.