Skip to content

An editorial team and a community building with AI

← All articles

News · GitHub · 2 min read

Event date: · Published:

GitHub Copilot Fusion became generally available, choosing and orchestrating models automatically

On October 28, 2026, GitHub made Copilot Fusion generally available, combining multi-model orchestration with auto model selection so users chose an effort level instead of a model.

Archive entry: this covers an event from October 28, 2026, and we published it on November 8, 2026. The details come from sources dated at the time and may have changed since.

On October 28, 2026, GitHub made GitHub Copilot Fusion generally available. Fusion combined Project HydraFusion's multi-model orchestration with Copilot auto model selection, so users chose an effort level instead of a model. GitHub described it in a changelog entry on routing and orchestrating models with Copilot Fusion.

What changed

  • Effort levels were Fusion Standard, Fusion High and Fusion Max, and Fusion orchestrated models through Single, Cascade and Critique workflows.
  • In the Artificial Analysis Coding Agent Index, Copilot Fusion Max scored 60, within two points of Claude Code with Opus 5.5 Max.
  • GitHub said Fusion Max reached that score at about half the average API cost per task, using about half as many tokens and about 40% less time per task.
  • Fusion used only models enabled in enterprise and organization model policies, admins could set Fusion as the default model, and users were billed for each model Fusion used.

What it meant for people building with AI

Copilot could choose models itself. By GitHub's numbers, Fusion Max came close to Claude Code's top setting at about half the cost. Because Fusion used only models enabled in model policies and each model it used was billed, admin-set policies still determined which models took part in a task.

Local model option

An experimental option let Fusion select local models: MAI Code 1.1 Flash through Windows ML on Windows, or MAI Code 1.1 Flash or Qwen3.8 Flash Next through an OpenAI-compatible server on Windows, macOS and Linux. It was enabled with the chat.copilot.localModel.enabled setting, and routing still happened in the cloud. GitHub covered it in a second changelog entry on local model selection.


Sources

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.

GitHub Copilot Fusion became generally available, choosing and orchestrating models automatically | vibecoding.inc