Microsoft unveils mai Code1.1Flash model, expands GitHub copilot's hybrid inference capabilities

Microsoft announced at its Windows and Surface launch event on October 7, 2026, that it will introduce local AI model support for GitHub Copilot by the end of October. This update allows developers to switch or automatically schedule between cloud and device-side models. To address memory and resource challenges in edge inference, Microsoft introduced the MAI Code1.1Flash Mixture of Experts (MoE) model, which has 137 billion parameters and 6.8 billion activated parameters. The model uses quantization and speculative decoding to reduce memory usage and improve response speed. It is integrated into the new Surface Laptop Ultra with NVIDIA RTX Spark hardware. Users can choose inference modes through GitHub Copilot CLI, the Copilot app, and Visual Studio Code. They can either let Copilot manage resources or specify local models like MAI Code1.1Flash via Windows ML providers or OpenAI endpoints. Microsoft also integrated its Executable Container (MXC) to ensure code execution security. This move marks the shift of AI coding assistants from cloud-only to a hybrid 'cloud + edge' architecture, simplifying local model deployment and offering efficient solutions for privacy and low-latency scenarios.