copilot added its first open-weight model — and enterprise teams have to opt in
Kimi K2.7 Code is the first open-weight model selectable in GitHub Copilot's model picker. it is off by default for Business and Enterprise — admins must flip the policy switch. the open-weight distinction matters more for auditing than for pricing.
GitHub added Kimi K2.7 Code to Copilot's model picker on July 1, 2026 — first for Pro, Pro+, and Max plans, then for Business and Enterprise on July 7. it is the first open-weight model to land in Copilot's five-lab roster.
the model is hosted by GitHub on Microsoft Azure and billed at provider list pricing under usage-based billing. Kimi K2.7 is from Moonshot AI — not Anthropic, OpenAI, Google, or Microsoft. the open-weight part means the model weights are public and inspectable.
open-weight is the story, not the cost
for most solo vibe coders, open-weight vs closed is academic — you pick whichever model gives the best completions for your use case. for enterprise buyers, it changes three things:
- auditability— security and compliance teams can evaluate what the model actually does, not just trust a vendor's attestation
- no lock-in— if GitHub's hosting terms change, you could self-host the same weights and get identical behavior
- reproducibility— weights don't change between API calls the way closed models can be silently updated. what passes in testing is what runs in production
pricing is still provider list pricing through GitHub, so the per-token cost isn't dramatically different from other selectable models today. the auditability angle is what should move enterprise adoption decisions.
the opt-in architecture for teams
for Copilot Business and Enterprise, Kimi K2.7 Code is disabled by default. an org or enterprise admin must go into Copilot policy settings and explicitly enable the Kimi K2.7 Code policy. if the policy stays off, the model stays out of the picker for everyone in the org — individual developers cannot self-provision it.
that opt-in architecture is the same pattern GitHub has used for other third-party models (Gemini, Mistral) as they came online. it gives legal and security teams a veto before any new lab's model touches your codebase, even if individual developers want access.
the practical question for teams: does your security team care about open-weight auditability enough to greenlight the policy switch? if yes, it's one admin toggle. if not, the existing claude-sonnet and gpt-class models cover most workflows without the policy conversation.