GitHub Copilot brings HydraFusion’s multi-model workflows to VS Code and the Copilot app

SNACK: 3-line summary

  • GitHub has expanded the HydraFusion research preview to VS Code and the GitHub Copilot app.
  • HydraFusion chooses between single-model, draft-and-escalate and draft-and-critique workflows.
  • Copilot Pro, Pro+, Business and Enterprise users can try it; Business and Enterprise administrators must enable preview features.

GitHub expanded the HydraFusion research preview on September 30, 2026, adding VS Code and the GitHub Copilot app alongside Copilot CLI. Developers can now try a workflow that decides when one model should handle a task and when drafting, escalation or a separate critique could help—all within those interfaces.

Official GitHub HydraFusion research-preview illustration published September 4, 2026
Official illustration from the September 4, 2026 HydraFusion research introduction, not a screenshot of the new VS Code support. Image: GitHub.

Snackgirls react

AIKO: A second pair of eyes that cannot edit the code—that’s an interesting division of labor. I’m curious whether a different model family notices mistakes the drafter misses.

Red: I’d try it on a stubborn bug first. If the task takes a while, those progress messages would give me something to follow instead of just waiting for the answer.

What happens after you submit a task

HydraFusion is a runtime orchestration system, not a single new AI model. It uses signals about reasoning, code generation, debugging and tool-use capabilities to choose an execution pattern expected to meet a quality bar.

  • Single: one selected model works directly on the task.
  • Cascade: an efficient model produces a draft. A quality gate either accepts it or escalates the task to a stronger model.
  • Critique: a model produces a draft, then a read-only critic from a different model family reviews it independently. The drafting model revises it once.

Copilot’s Auto option picks a model for each request. HydraFusion also selects the workflow and can coordinate multiple models within a turn. That does not mean every request uses several models or that the process guarantees passing tests.

How to enable it in VS Code or the app

  • In VS Code 1.140 or later, or VS Code Insiders, select HydraFusion in the Copilot Chat model picker. If it is missing, enable chat.copilot.hydraFusion.enabled.
  • Update the GitHub Copilot app to the latest version. Search Settings for HydraFusion, turn it on, then select it in the model picker.

The preview supports Copilot Pro, Pro+, Business and Enterprise. For Business and Enterprise users, an administrator must enable preview features before they can use it.

Clearer progress during longer tasks

The September 30 release also makes workflow stages more transparent, provides more frequent real-time progress messages and shows clearer status during long tasks. HydraFusion remains a research preview and is subject to change.

Sources and date: 2026-10-01 KST

Related hashtags
#GameSunakku #AI #GitHubCopilot #HydraFusion #VSCode #AICoding

Game Sunakku에서 더 알아보기

지금 구독하여 계속 읽고 전체 아카이브에 액세스하세요.

계속 읽기