GitHub Copilot Fusion adds selection of local models in experimental preview
GitHub announced an experimental preview of intelligent routing for local models in GitHub Copilot Fusion. Copilot Fusion can now choose between supported on-device models and cloud models.

Why it matters
Builders using Copilot Fusion may see requests routed to on-device models instead of cloud models, which can affect latency, data locality, and cost. The feature is in experimental preview, so behavior may change.
Who should care
Developers using GitHub Copilot, engineering teams with data residency or latency requirements, tooling builders integrating with Copilot
What you can do
Enable the experimental preview if available to your account and test which tasks are routed to local versus cloud models before relying on it in production.

