Microsoft adds HydraFusion multi-model AI to VS Code

HydraFusion is not a new model but a way of routing work between existing ones. Microsoft describes it as a research project, detailed in a GitHub blog titled "Project HydraFusion: Frontier quality via multi-model orchestration." For each task the user gives, it uses capability signals from AI reasoning, code generation, debugging, and tool use to choose the most efficient execution pattern that still meets a quality bar — the stated aim being high quality at lower cost.

The three patterns differ in how much verification they add. Cascade has an efficient model draft a solution and a quality-check model verify it before deciding whether to escalate to a stronger model. Critique has one model draft, a model from a different family review it independently, and the drafting model revise once based on that feedback. This is a different approach from Copilot's Auto mode, which picks a model per request rather than combining models within a task's workflow.

Microsoft says that when new AI models become available in GitHub Copilot, they will be evaluated and folded into HydraFusion's model pool. Whether the approach delivers on its cost-and-quality promise likely hinges on how well those routing decisions hold up across real coding tasks — and on how many Copilot Business and Enterprise administrators turn the preview on.