HydraFusion updates and new Max option
By Steven Van ·
Both options remain in research preview across VS Code, the GitHub Copilot app and CLI on Pro, Pro+, Business and Enterprise plans.
Copilot’s HydraFusion now sets reasoning effort for each stage of a task and has a new Max option for more demanding work. GitHub has also updated its model pool and the workflows that draft, judge and revise a response.
Models and reasoning effort
HydraFusion and HydraFusion Max now draw from a pool that includes Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. The models used vary by task and workflow, and the pool can change during the preview.
Previously, HydraFusion used each model’s default reasoning effort. It now uses medium effort for drafts and high effort for reviews. HydraFusion Max applies higher reasoning effort to tasks where depth and thoroughness matter, while still considering token efficiency.
Drafting and review
Selecting HydraFusion in the model picker lets it choose an execution pattern, models and reasoning effort for each prompt. It returns a single final response, with progress visible as each step runs.
- Single: One model completes the task.
- Cascade: An efficient model drafts a solution. A judge inspects it and decides whether to accept it or escalate to a stronger model.
- Critique: Multiple models independently review a draft in parallel. Their feedback is combined, and the original model revises its response once. The updated workflow relies more on tools to inspect the workspace directly.
Access and billing
Both options remain in research preview across VS Code, the GitHub Copilot app and Copilot CLI, on Copilot Pro, Pro+, Business and Enterprise plans. Access depends on preview settings and which models the plan and organization policies allow.
In VS Code 1.140 or later, or VS Code Insiders, enable chat.copilot.hydraFusion.enabled. In the updated Copilot app, turn on HydraFusion under Settings, Experimental. In the latest Copilot CLI, enable experimental features, then select either option through /model.
The HydraFusion documentation says each model pass is billed at that model’s standard rate, with no separate orchestration charge. The Auto discount does not apply, and review passes can take longer and consume more AI credits. The preview has no SLA and is not intended for production workloads.
