GitHub Copilot Fusion: choose the effort, let it pick the models

GitHub made Copilot Fusion generally available on October 28, 2026, bringing together Project HydraFusion’s multi-model orchestration and Copilot’s automatic model selection. You choose an effort level for the task, and Fusion handles which models to use and how they work together, returning one final response.

A task can still use just one model

The three selectable levels give Fusion different priorities. Standard selects one model per task for everyday coding. High targets more challenging coding and reasoning, while Max focuses on the most demanding problems. High and Max can add model calls when the task warrants them.

Fusion chooses among three execution patterns. In Single, one model completes the task. In Cascade, an efficient model produces a draft, then a judge decides whether to accept it or escalate to a stronger model. In Critique, other models review an initial draft, and the original model revises it using their feedback.

The pattern can change between prompts in the same session. A straightforward request does not automatically trigger a round of multi-model review.

Where Fusion appears—and who controls its model pool

Fusion is available across all Copilot plans in Visual Studio Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs. Update your client, open the model picker, select Copilot Fusion and choose an effort level.

Support across all plans does not mean every account gets every effort level or model. Fusion can use only models allowed by your subscription and enterprise or organization policies. If none of its models are available to you, Fusion will not appear in the picker.

Administrators can restrict Fusion’s model pool, including excluding a model from Fusion without disabling it for everyone. Enterprise managed settings also let them make Fusion the default at a chosen effort level.

Extra review calls can mean extra credits

Each model Fusion uses is billed at that model’s standard rate, with no separate Fusion surcharge. Through November 30, 2026, paid Copilot plans receive a 30% discount on model costs when using Fusion. The offer applies to model usage, not the subscription fee.

High and Max can consume more AI credits than Standard because they may call additional models. The discount does not guarantee that a higher-effort task will cost less overall.

A discarded draft can leave file edits behind

If Fusion discards an intermediate draft, any changes that draft already made in your workspace are not automatically undone. Review the diff before committing: the final response alone does not tell you which file edits remain.