Introduction to Intelligent Local Model Routing

GitHub has announced the release of intelligent routing for local models into experimental preview. With this update, GitHub Copilot Fusion can now choose between supported on-device models and cloud models. This capability enables intelligent local model routing across the GitHub Copilot CLI, GitHub Copilot app, and Visual Studio Code, letting users utilize their device’s compute without having to switch models manually for every request.

For a deeper dive into this feature, developers can review the details shared in our original announcement on October 7th.

Setup and Configuration Across Platforms

On supported Windows devices, users can download and configure MAI Code 1.1 Flash through the Windows ML provider. In the Copilot CLI, developers must enable experimental mode and open /model to follow the setup prompts. Detailed configuration instructions are available in the documentation for Adding LLM models to GitHub Copilot CLI.

In the Copilot app, users need to configure the Windows ML provider under Settings > Model providers, enable Local model auto routing under Settings > Experiments, and start a new session. Windows ML local models are enabled by default on Windows. For app-specific setup instructions, consult the documentation covering Adding LLM models to the GitHub Copilot app. In VS Code, users simply need to enable chat.copilot.localModel.enabled.

Selection and Managed Experience Details

Once the model is ready, users can select Standard in Copilot Fusion to let Copilot choose between local and cloud models. Alternatively, users can select the local model directly for their session.

While the managed download experience is strictly Windows-only, the Copilot CLI's local Copilot Fusion routing also functions with manually configured models on Windows, macOS, and Linux. Developers can configure MAI Code 1.1 Flash or Qwen3.8 Flash Next through a local OpenAI-compatible server. Users manage the model download and inference runtime directly, provided their device has sufficient resources to run it. Users must configure the model as a supported local-model provider, enable experimental local routing, and select Standard in Copilot Fusion to let Copilot choose between eligible local and cloud models.

BYOK Integration and System Limitations

This release builds on the recent local-model discovery improvements in Copilot CLI. However, adding or discovering a model through bring your own key (BYOK) configuration does not automatically make it eligible for Copilot Fusion.

The Copilot app also supports connecting and directly selecting compatible local models through BYOK, though its intelligent routing currently supports managed Windows ML models only.

GitHub notes that local inference does not render a session offline. Copilot Fusion continues to make routing decisions in the cloud, and local models may feature different safety protections compared to cloud models provided directly by Copilot.

For complete information on routing behavior, users can refer to our docs on intelligent routing to local models.

Sources

• GitHub ChangelogGitHub Copilot Fusion now supports selection of local models

Continue chronologically

Newer storyGitHub Copilot Updates in VS Code Bring New Agent FeaturesOctober 28, 2026 · 3 min

You are readingGitHub Copilot Fusion Adds Support for Local Models

Older storyMicrosoft Introduces Copilot Harness for Multi-Model AIOctober 28, 2026 · 3 min

Related stories

More on this topic

Apps & Software

GitHub Tightens Enterprise Copilot Settings Enforcement

GitHub is giving enterprise administrators stricter control over Copilot policies with fresh settings checks and minimum client versions.

October 13, 2026 · 4 min

Artificial Intelligence

GitHub Copilot Assisted Permissions Generally Available

GitHub Copilot assisted permissions are now generally available in Visual Studio Code, the GitHub Copilot app, and the GitHub Copilot CLI.

October 27, 2026 · 4 min

Artificial Intelligence

GitHub Fixes Copilot Metrics Bug Affecting Agent Activity

GitHub is rolling out an IDE fix to restore missing agent activity and lines of code in Copilot usage metrics following a recent SDK transition issue.

October 7, 2026 · 4 min

Artificial Intelligence

GitHub Copilot adds model tiers, reviews and Sentry

GitHub Copilot’s latest weekly release adds model-selection tiers, code review improvements, Sentry integration and new agent controls.

September 19, 2026 · 4 min