HydraFusion stops making you pick a model (and adds a “Max” mode)

In plain English

Stop guessing which model to pick: HydraFusion now draws on a bigger pool for you, and Max adds a setting for the hardest jobs.

HydraFusion is a Copilot option that runs several models on one task and combines their work into a single answer, so you don’t have to choose a model yourself. It is still a research preview.

What changed

• The model pool now includes Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Which models a task uses depends on the workflow.
• Reasoning effort is set per step: medium for first drafts, high for review.
• Cascade can add a judging step that checks the first attempt and asks for a repair only when needed. Critique now inspects your workspace with tools and runs several critic models in parallel.
• HydraFusion Max uses the same approach with higher reasoning effort for demanding tasks, while still keeping token use in mind.

Where you get it

In VS Code, the GitHub Copilot app and Copilot CLI, on Copilot Pro, Pro+, Business and Enterprise plans with experimental or preview features turned on. Your organization’s model policies still apply.

Which one to use

Plain HydraFusion suits everyday agent tasks that balance speed, cost and quality. Max is for tasks that need the deepest reasoning and the most thorough checking.