GitHub previews HydraFusion, which assembles a multi-model workflow for each Copilot CLI task

GitHub has released Project HydraFusion, a research preview that treats model selection as a per-request optimization rather than a fixed setting. Instead of routing a prompt to a single model, HydraFusion builds an execution plan for each task. It can draft with one model, have a second model critique the draft, or escalate to a stronger model when a quality gate rejects the first attempt. The models come from multiple providers, but the developer selects HydraFusion once, the same way they would choose any other model. Availability is narrow. HydraFusion is live for users on all GitHub Copilot plans, but only inside GitHub Copilot CLI. There are no open weights and no self-hosted option. To enable it, users run the update command, turn on experimental features, and choose HydraFusion (Research Preview) from the model list. Billing is per token consumed by whichever models the workflow invokes, at each model's standard rate, so a task that escalates through several models costs more than a single-model run. The feature builds on Auto model selection, which GitHub shipped earlier in 2026 to match a task to one best-suited model. HydraFusion extends that idea by reading capability signals for reasoning, code generation, and debugging and selecting an entire workflow rather than one model.