From the source
GitHub is shipping intelligent routing for local models in GitHub Copilot Fusion as an opt-in experimental preview, letting Copilot choose between supported on-device models and cloud models.
- The routing works in GitHub Copilot CLI, the GitHub Copilot app and Visual Studio Code, so developers can use their device's compute without switching models manually for each request.
- On supported Windows devices, MAI Code 1.1 Flash can be downloaded and configured through the Windows ML provider, and Windows ML local models is enabled by default on Windows.
- The managed download experience is Windows-only, but Copilot CLI's local Copilot Fusion routing also works with manually configured models on Windows, macOS and Linux.
- Users can configure MAI Code 1.1 Flash or Qwen3.8 Flash Next through a local
OpenAI-compatible server, managing the download and inference runtime themselves, and then select Standard in Copilot Fusion to let Copilot choose between eligible local and cloud models.
- GitHub warns that local inference does not make a session offline, since Copilot Fusion still makes routing decisions in the cloud, and local models may have different safety protections from Copilot-provided cloud models.
- Adding or discovering a model through bring your own key configuration does not automatically make it eligible for Copilot Fusion, and the Copilot app's intelligent routing currently supports managed Windows ML models only.
