Microsoft opened Build 2026 in San Francisco with the biggest shift in GitHub Copilot's history: Project Polaris, a Microsoft-built coding model that will replace GPT-4 Turbo as the default backbone for all Copilot subscribers beginning August 2026.
Project Polaris was designed specifically for software engineering workloads — code generation, multi-file refactoring, test writing, code review, documentation, and dependency analysis — rather than being adapted from a general-purpose LLM. Microsoft says the model runs natively on its Maia 200 AI accelerators inside Azure, giving the company end-to-end control of both the silicon and the model serving the world's most-used AI coding assistant.
The announcement marks the most visible step yet in Microsoft's gradual decoupling from OpenAI's frontier model lineup for internal products. While OpenAI models will still be selectable inside Copilot, Polaris becoming the default model touches tens of millions of developer seats across GitHub Enterprise, Copilot Workspace, and Visual Studio.
Microsoft paired Polaris with two adjacent launches that frame the company's broader agent push: Copilot Workspace went generally available for GitHub Enterprise, and Agent Mode became the default in Office 365 Copilot across Word, Excel, and PowerPoint. The Build keynote also confirmed commercial availability of MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — Microsoft's own multimodal stack — alongside Foundry Local for on-device inference on Windows, macOS, and Linux.
For developers, the practical change lands in two phases: Polaris becomes opt-in this month for Copilot Pro and Business tiers, then takes over as default in August. Microsoft says enterprises that pin a specific model can keep doing so, but unpinned tenants will silently migrate. Independent benchmarks have not yet been published.
Source: [BuildFastWithAI](https://www.buildfastwithai.com/blogs/ai-news-today-june-2-2026)