GitHub launches hydrafusion: a new tool for dynamic model orchestration in copilot cli

GitHub has introduced Project HydraFusion, a research preview that dynamically selects and orchestrates multiple models to optimize coding workflows. Unlike traditional approaches, HydraFusion constructs an execution plan for each request, allowing for different models to draft, critique, or refine code based on quality checks. The tool is available to users on all GitHub Copilot plans via the Copilot CLI, with billing based on token usage per model. HydraFusion builds on GitHub’s earlier Auto model selection feature, taking a step further by treating workflow selection as an optimization problem. It evaluates capabilities like reasoning and debugging to choose the least complex workflow that meets quality standards. The tool was tested on three coding benchmarks using Claude Opus 5 and GPT-5.6 Sol as baselines, with results relative to Opus 5. GitHub also highlighted its internal CheckpointBench dataset, which is based on real Copilot sessions and public commits.