Microsoft to unveil homegrown coding model at Build

TL;DR:

  • Microsoft will reportedly unveil a suite of in-house AI models at its Build developer conference next week, including a coding model intended to boost GitHub Copilot.
  • The line-up is said to also cover transcription, reasoning, speech and images, according to the Information; Microsoft declined to comment and shares rose nearly 3% on the report.
  • The push reflects Microsoft’s effort to reduce dependence on OpenAI, whose partnership terms have been renegotiated, as rivals’ coding tools — notably Anthropic’s Claude Code — have pulled ahead.

Microsoft appears ready to stake out independence in the one AI category it arguably pioneered commercially: AI-assisted coding. The reported plan to debut homegrown models at Build, led by a coding model for GitHub Copilot, signals a strategic shift from integrating partners’ models to building its own across transcription, reasoning, speech and imaging.

The competitive pressure behind the move

The context is a partnership in transition. Microsoft has leaned heavily on models from OpenAI, Anthropic and Google to power GitHub Copilot, but the once-central OpenAI relationship has been reworked in recent months to reduce mutual reliance. Meanwhile Copilot, an early leader, has been overtaken: products such as Anthropic’s Claude Code have risen to the top of AI-assisted coding, a reversal that helps explain the urgency.

Microsoft is also reportedly eyeing AI start-up acquisitions to add talent and hit its stated goal of building a cutting-edge model by next year. Investor sentiment has cooled this year amid questions about whether Microsoft’s early AI lead is durable as the OpenAI partnership unwinds and Google and Amazon report strong progress — making Build a moment to reassure markets as much as developers.

Looking forward

For UK developers and the enterprises that employ them, more model competition in coding tools should mean better options and pricing leverage, reducing single-vendor lock-in. The open question is quality: Microsoft must show its homegrown coding model can close the gap with the tools that overtook Copilot, not merely replace a partner’s model with its own. Build will be the first real test of whether Microsoft can compete on capability, not just distribution.