Microsoft Brings Local AI Inference to GitHub Copilot This Month

Microsoft plans to add local AI model inference to GitHub Copilot before the end of October, giving developers more control over where coding-assistant requests are processed.

The hybrid architecture will let users switch automatically or manually between cloud-based models and models running on their devices. A separate mode will force processing through a local model.

Developers will be able to set their preferences in GitHub Copilot CLI, the Copilot app and Visual Studio Code. GitHub Copilot generates code completions, explanations and refactoring suggestions based on the surrounding code.

The local-model feature will support selected providers, models and endpoints. Windows ML users will be able to choose MAI Code 1.1 Flash, while developers using OpenAI-compatible local endpoints can select models exposed by those endpoints.