MICROSOFT FIRES OPENAI FROM GITHUB COPILOT AND BUILDS ITS OWN MODEL — THE DIVORCE THAT CHANGES EVERYTHING At its annual Build conference in San Francisco on Tuesday, Microsoft announced Project Polaris, an in-house AI coding model that will replace OpenAI GPT-4 Turbo as the default engine inside GitHub Copilot starting August 2026. The move signals the end of one of the most commercially significant partnerships in tech history. Microsoft invested billions in OpenAI, baked its models into every major product, and built a developer ecosystem on top of them. Now it is building a model of its own. Polaris is a mixture-of-experts architecture with specialized sub-modules tuned for different programming languages. Microsoft says it outperforms GPT-4 Turbo on standard coding benchmarks, particularly in lower-resource languages like Rust and Haskell, and runs on Microsoft own Maia AI accelerators inside Azure, cutting per-inference costs and stripping out the OpenAI line item entirely. The strategic logic is clean: Microsoft now controls the model, the hardware it runs on, the developer tools, and the distribution. Build also brought the Windows Agent Framework into open source and moved three Copilot Studio capabilities into general availability: computer-using agents, agent-to-agent communication, and real-time voice. Windows is no longer just an operating system. Microsoft is positioning it as a runtime for autonomous AI agents that plan and execute work end-to-end. For OpenAI, which generates enormous revenue from the Microsoft relationship, this is not a minor product update. It is a slow-motion renegotiation of the entire partnership. Keywords: Microsoft Project Polaris, GitHub Copilot AI model, Microsoft Build 2026, Windows AI agents