Where the preview is available
GitHub’s 14 October announcement adds HydraFusion Max alongside the existing HydraFusion option in VS Code, Copilot CLI and the Copilot app. It names Pro, Pro+, Business and Enterprise plans, with preview features enabled and organization model policies still applying.
GitHub describes an updated pool including Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. The selected models vary with the workflow; this is not a promise that each request uses every model.
Multiple passes change the cost calculation
The usage documentation says each underlying model is billed at its standard rate. There is no separate HydraFusion fee, and the Auto selection discount does not apply. Review passes can increase the AI credits consumed by a task; no fixed task price is established here.
The documented patterns are one model pass, a draft checked for possible escalation, or a draft with parallel critiques and a revision. Max can use higher reasoning effort. GitHub’s capability and efficiency descriptions remain vendor claims.
Workspace changes need their own review
GitHub labels the feature a research preview without an SLA, outside its intended production workload scope. Its documentation also says discarding a draft does not automatically undo workspace file edits. Check the diff before committing.
Plan and model-policy restrictions can hide the picker option. The documentation gives client-specific setup steps; account eligibility and regional access were not independently tested.
Keep a trial comparable
Record the client, plan, permitted models, execution pattern, visible usage and accepted result. Use the coding task worksheet to keep those conditions alongside the output.
See the Copilot profile and Cursor/Copilot comparison for separately dated context. This source check establishes no new winner, score or numeric price.