GitHub Copilot Fusion offers experimental routing between local and cloud models
GitHub has made experimental automatic selection between supported local and cloud models available. The feature requires users to opt in, and its availability depends on the application and operating system. Routing decisions continue to be made in the cloud.
GitHub has made experimental routing of requests between supported local and cloud models available in GitHub Copilot Fusion. The feature requires users to opt in and is available in GitHub Copilot CLI, GitHub Copilot app and Visual Studio Code. Once a supported local model is configured, the Standard option enables automatic selection without manual switching for each request.
On supported Windows devices, MAI Code 1.1 Flash can be downloaded and configured through Windows ML. GitHub Copilot CLI also supports manual configuration of MAI Code 1.1 Flash or Qwen3.8 Flash Next through a local server with an interface compatible with the API from OpenAI, on Windows, macOS and Linux. In this case, users are responsible for downloading the model, providing its runtime environment and ensuring sufficient hardware resources.
Automatic routing in GitHub Copilot app currently supports only models managed through Windows ML. Compatible local models connected through BYOK can be selected directly in this application; however, simply adding or discovering a model through BYOK does not guarantee its eligibility for automatic routing. Routing decisions are made in the cloud, so local inference does not ensure offline operation. According to GitHub, the security safeguards of local models may also differ from those of the cloud models provided.
Why it matters
Developers can include the computing resources of their own devices in automatic model selection. However, the configuration method matters when choosing a working environment: automatic routing supports manually configured local models in GitHub Copilot CLI on three operating systems, while GitHub Copilot app limits this feature to models managed through Windows ML.
Two audiences, two different impacts
What this means
For individuals
When working in GitHub Copilot CLI on Windows, macOS and Linux, you can include a supported local model in automatic selection and reduce manual switching between models.
For a business
When deploying local inference in a business, you must account for cloud-based routing decisions and possible differences in model security safeguards. The feature therefore does not, by itself, ensure offline operation.
Risks and complianceCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.