GitHub confirmed on 7 July 2026 that Kimi K2.7 Code — an open-weight model from Moonshot AI, hosted by GitHub on Microsoft Azure — is now generally available inside Copilot for Business and Enterprise customers, having rolled out to Pro plans a week earlier. It’s the first open-weight model to appear as a selectable option in Copilot’s model picker, sitting alongside the closed frontier models from Anthropic, OpenAI, and Google. For Business and Enterprise plans it’s opt-in: administrators have to enable it manually before developers can select it.
This matters less for the specific model and more for what it signals about how the model layer inside AI coding tools is evolving. Copilot, Cursor, and others have spent the past year turning themselves into multi-model platforms rather than single-model products — letting teams pick a model per task, per cost tier, or per data-handling requirement. An open-weight option inside that mix gives engineering teams with strict data-residency or self-hosting requirements a path that wasn’t previously available inside Copilot at all, even if most teams will stick with the default closed models for now.
So what
If your organisation has data-governance or self-hosting requirements that have kept AI coding assistants off the table, this is worth a fresh look — model choice inside these tools is no longer fixed, and an open-weight option changes the calculus for regulated or security-conscious teams. For most commissioning conversations the practical takeaway is simpler: which model runs inside your AI coding tool is now a genuine, changeable decision, not a vendor lock-in. Our custom software development work includes advising on tooling choices as part of a build — get in touch if you want a second opinion.