Microsoft eyes in-house Copilot coding model

If this lands, it is less about model novelty than Microsoft trying to regain control of Copilot’s margins, roadmap, and developer mindshare. The bigger question is whether an in-house model can close the quality gap fast enough to matter in a market now benchmarked against Cursor, Claude Code, and OpenAI’s coding stack. Reuters says the new model is meant to boost GitHub Copilot usage, which implies Microsoft sees a competitive problem, not just a platform opportunity. GitHub’s own product page says Copilot already uses models from GitHub, OpenAI, and Microsoft, so this move looks like an attempt to shift more of the value chain in-house. Build is a logical venue because Microsoft can pair the model story with IDE, CLI, GitHub, and Azure distribution in one shot. Even a slightly weaker model could still win enterprise share if Microsoft prices it aggressively and bundles it cleanly across Copilot plans. Developers will care less about the “homegrown” label than whether it improves code quality, latency, and agent reliability on real repos.