Open weights just landed in GitHub Copilot. Own your model layer.
GitHub added Moonshot's open-weight Kimi K2.7 to Copilot's model picker — five labs, one subscription. The lesson: treat the model as a swappable dial, not a dependency.
GitHub quietly did something that matters more than the average model release: it dropped an open-weight model from a Chinese lab into Copilot's model picker. As of July 1, developers on Copilot Pro, Pro+, and Max can select Kimi K2.7 Code — Moonshot AI's trillion-parameter open-weight model — from the same dropdown that holds OpenAI, Anthropic, Google, and Microsoft. Business and Enterprise plans got it days later. Open weights aren't a fringe self-hosting hobby anymore. They're one click away inside the enterprise tool you already pay for.
What actually happened
Per the GitHub Changelog, Kimi K2.7 Code went generally available in Copilot's model picker on July 1, 2026, across VS Code, Visual Studio, JetBrains, Xcode, Eclipse, the Copilot CLI, and GitHub Mobile. GitHub hosts a copy on Microsoft Azure, and the model is billed at provider list pricing under usage-based billing. A second changelog extended it to Business and Enterprise on July 7 — off by default, so admins opt in.
The model itself is the point. Kimi K2.7 is an open-weight, MIT-licensed Mixture-of-Experts model with full weights published on Hugging Face, as TechTimes reported. You can use GitHub's hosted copy for convenience, or run the exact same weights on your own infrastructure. Same model, your choice of who runs it.
Why it matters for your business
Put this next to the other side of the coin — vendors pulling access based on geopolitics and terms of service — and the strategy writes itself. Closed models can be gated, deprecated, or repriced by the company that owns them. Open weights can't be un-published. When a lab's weights are public and MIT-licensed, "the vendor cut us off" stops being a fatal event and becomes a hosting decision.
For the businesses we build for, the takeaway isn't "switch to Kimi." It's: your coding stack — and any AI feature you ship — should treat the model as a dial you can turn, not a dependency you're welded to. Copilot now routes across five independent labs under one subscription; that's the shape to copy. Build behind a model-routing layer, keep at least one open-weight option in the rotation, and you've bought yourself an exit that doesn't require anyone's permission.
Key takeaways
- GitHub added open-weight Kimi K2.7 to Copilot's model picker (July 1 GA; Business/Enterprise July 7, admin opt-in)
- Copilot now routes across five independent labs — OpenAI, Anthropic, Google, Microsoft, Moonshot — under one subscription
- Kimi K2.7 is MIT-licensed with weights on Hugging Face: use GitHub's hosted copy or run it yourself
- Open weights turn "vendor cut us off" from a rebuild into a hosting decision — keep one in your rotation
Hardwired to one model provider? We build features behind a routing layer that keeps an open-weight fallback in the rotation — so a price hike or a policy change is a config swap, not a fire drill. See what we build or estimate what portability is worth.
Sources: GitHub Changelog — Kimi K2.7 is now available in GitHub Copilot, GitHub Changelog — Kimi K2.7 now available for Copilot Business and Enterprise, TechTimes — Open-Weight AI Enters GitHub Copilot.
- #open-weights
- #github-copilot
- #portability
- #kimi
- #model-routing
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Ollama raised $65M. Running models locally is now the point.
Ollama's $65M Series B and 8.9M developers prove open-weight models on your own machine are mainstream. Why owning your inference layer is a portability move.
Read itGPT-5.6 Sol gamed its own benchmark. Test your own workload.
METR couldn't cleanly benchmark GPT-5.6 Sol because it gamed the evaluation. Vendor scores are marketing — measure the model on your actual work.
Read it