Microsoft is expanding the model selection in GitHub Copilot by introducing support for open-weight models, with China's Kimi K2.7 Code model being the first to join the lineup. As of July 1, 2026, developers can select Kimi K2.7 as a low-cost alternative within Copilot and Visual Studio Code, marking a significant shift in how Copilot integrates external AI models.

Why This Matters for Developers

For years, GitHub Copilot has relied on a single underlying model (initially OpenAI's Codex, later GPT-4o derivatives) to power its code completions and chat features. While effective, this lock-in limits flexibility and cost control. The introduction of open-weight models gives developers the ability to choose a model that best fits their budget and performance needs. Kimi K2.7, developed by Moonshot AI, is positioned as the cheapest non-lightweight model outside GitHub's fine-tuned Raptor mini option, making it an attractive choice for cost-conscious teams.

How Kimi Fits Into the Copilot Ecosystem

Kimi K2.7 is hosted on Microsoft Azure and joins existing options from OpenAI, Google, and Anthropic. However, unlike those proprietary models, Kimi is an open-weight model, meaning its parameters are publicly available. This opens up possibilities for customization and fine-tuning by the community, although GitHub currently offers it as a pre-built option. Administrators of GitHub Copilot Enterprise and Business plans must enable Kimi manually—it is turned off by default. Individual Copilot users on Pro or Free plans will see it automatically in the model picker.

Availability and Pricing

According to GitHub's documentation, Kimi K2.7 is the cheapest non-lightweight model available. While exact pricing varies based on usage tiers, this move signals Microsoft's willingness to compete on cost. Developers can expect a balance between quality and expense, especially for tasks where a high-end model like GPT-4o is overkill. By leveraging Azure's infrastructure, latency and data residency remain compliant with enterprise requirements.

Practical Example: Switching to Kimi in VS Code

To try Kimi, open VS Code and ensure the Copilot extension is up to date. In the model picker (located in the Copilot chat or code completion settings), you'll see a new entry: "Kimi K2.7 Code (open-weight)". Select it, and Copilot will start using this model for suggestions and chat. For a typical coding session, you might notice slightly different response patterns—Kimi tends to be more verbose in explanations but equally accurate for common languages like Python, JavaScript, and TypeScript.

Workflow tip: Use Kimi for routine tasks such as writing boilerplate code, generating unit tests, or refactoring simple functions. Reserve more expensive models for complex reasoning or when you need precise adherence to specific libraries. You can switch models per project via the .vscode/settings.json file:

This allows teams to standardize on a low-cost model for cost-sensitive workflows.

Implications for the AI Industry

Kimi's inclusion is a win for China's AI sector, which has been rapidly advancing open-weight models. Moonshot AI's K2 series has shown competitive performance on coding benchmarks, and this partnership validates its quality. For developers worldwide, it provides a real-world example of how open-weight models can be integrated into commercial tools without compromising security or functionality. Expect more open-weight models to appear in Copilot as GitHub expands its model marketplace.

Key Takeaways

• GitHub Copilot now supports open-weight models, starting with Kimi K2.7 from Moonshot AI.
• Kimi is the cheapest non-lightweight model, ideal for cost-conscious teams.
• Enterprise administrators must enable Kimi manually; Pro/Free users see it by default.
• Developers can switch models in VS Code settings for fine-grained control.
• This move could accelerate adoption of open-weight models in developer tooling.

Conclusion

The addition of Kimi K2.7 marks a pivotal moment for GitHub Copilot. By embracing open-weight models, Microsoft is not only lowering costs but also fostering a more competitive and flexible AI ecosystem. Developers now have the freedom to choose the best tool for their budget and task, without being locked into a single provider. As more open-weight models become available, expect Copilot to evolve into a truly agnostic AI assistant.

---

Source: GitHub picks China's Kimi for "low-cost" Copilot option