GitHub updated HydraFusion in GitHub Copilot on October 14, 2026, and introduced a Max tier. Both remain in research preview and are available across VS Code, the GitHub Copilot app and Copilot CLI. The announcement page carries a date without a time of day. Read the changelog entry.
What HydraFusion is
HydraFusion orchestrates multiple models to produce a single response, combining their strengths so the user does not have to pick one model per task. The models used vary with the orchestration workflow.
What changed
- Model pool: HydraFusion and HydraFusion Max now draw on Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna.
- Per-step reasoning effort: instead of each model's default effort, the workflows explicitly tailor effort to each stage — medium effort for initial drafts and high effort for review.
- Cascade workflow: can now add a judging step that inspects the initial result and requests repair only when needed, rather than always escalating to a stronger model.
- Critique workflow: relies more on tools that inspect the workspace directly instead of text summaries, and runs critic models in parallel.
HydraFusion Max
Max uses the same architecture and workflow techniques but is tuned for frontier capability on more demanding tasks, applying higher reasoning efforts where depth matters most while still considering token efficiency. The regular tier is described as a balanced option with moderate token savings for agentic tasks that balance speed, cost and performance.
The GitHub Copilot profile lists official entry points; recent model additions are covered in our reports on Gemini 4 Argon and Claude Haiku 5.5 in Copilot.
