Microsoft announced on October 8, 2026, that GitHub Copilot will soon support on-device AI coding on Windows PCs, combining local edge models and cloud-scale infrastructure.
Local Models and Orchestration
Microsoft announced that GitHub Copilot will soon determine when a task should be handled by on-device intelligence or cloud-scale models. The system automatically coordinates local and cloud inference behind the scenes for developers using NVIDIA RTX Spark Windows PCs, such as the Surface Laptop Ultra.
This setup uses a local version of MAI Code 1.1 Flash, a coding-optimized mixture-of-experts model featuring 137 billion total and 6.8 billion active parameters. The company stated that the on-device work applies quantization and speculative decoding to reduce the model footprint and improve responsiveness.
Sandboxed Tool Execution
To help secure agentic coding sessions, GitHub Copilot utilizes Microsoft Execution Containers (MXC). According to Microsoft, MXC is an open-source library that translates policy into native operating-system controls across different platforms.
On Windows, GitHub Copilot uses the BaseContainer tier of the ProcessContainer backend, while using Seatbelt on macOS and bubblewrap on Linux. These local backends apply policies to processes and services launched by the agent, controlling access to files, networks, credentials, and system capabilities.
Model Selection and Usage
GitHub Copilot is adding two ways to use local models across the GitHub Copilot CLI, Copilot app, and VS Code. Developers can let Copilot's Auto orchestration choose when to use local or cloud inference across multi-turn sessions.
Alternatively, explicit local-model selection supports workflows that need a specific provider or endpoint. Developers can select MAI Code 1.1 Flash through the Windows ML provider or connect GitHub Copilot to OpenAI-compatible local endpoints.
Key facts and where they come from
- GitHub Copilot is introducing automated routing between on-device intelligence and cloud-scale models.
Coming by the end of the month, GitHub Copilot will determine when a task is best handled by on-device intelligence and when it should leverage cloud-scale models.
- Microsoft AI developed a local version of MAI Code 1.1 Flash with 137 billion total and 6.8 billion active parameters.
Microsoft AI developed a local version of MAI Code 1.1 Flash, a coding-optimized mixture-of-experts model with 137 billion total and 6.8 billion active parameters.
- Surface Laptop Ultra features NVIDIA RTX Spark with up to 128 GB of unified memory and up to 1 petaflop of AI compute.
Surface Laptop Ultra is built around NVIDIA RTX Spark, with up to 128 GB of unified memory and up to 1 petaflop of AI compute.
- GitHub Copilot uses Microsoft Execution Containers (MXC) to apply policies and sandbox shell commands and tools.
GitHub Copilot uses Microsoft Execution Containers, or MXC, an open-source library from the Windows team that translates policy into native operating-system controls.