GitHub announced on September 30, 2026 that its HydraFusion research preview is now available in Visual Studio Code and the GitHub Copilot app, expanding beyond Copilot CLI. HydraFusion appears in the model picker, but GitHub emphasizes that it is not a single model. It is a workflow system that coordinates multiple models.
The system treats workflow selection as an optimization problem. It uses capability signals for reasoning, code generation, debugging, and tool use to choose an execution pattern that is efficient while meeting a quality bar. Single sends the task to one selected model. Cascade lets an efficient model draft first and then asks a quality gate whether to accept the result or escalate. Critique has one model draft, an independent read-only critic from another model family review it, and the drafting model revise once.
That is different from simply using Auto to select a model. GitHub says Auto chooses a model for each request, while HydraFusion considers both workflow selection and multi-model coordination within a turn. For coding agents, the unit of decision is moving from a model name to a task path: straightforward work can go directly to one model, while uncertain or higher-risk work can add a quality gate or independent review.
The release also improves visibility. Users can see more clearly what HydraFusion is doing at each step, receive more frequent real-time progress updates, and get clearer status during long tasks. These details are not headline model capabilities, but they matter for agent workflows. If users cannot tell whether a system is waiting, switching, retrying, or failing, they cannot decide when to intervene.
HydraFusion is available to Copilot Pro, Pro+, Business, and Enterprise users. Business and Enterprise administrators must enable preview features. It remains a research preview and may change. That limits what can be concluded from the announcement: it describes the product shape and availability, but this update does not provide an independent benchmark proving that every task will improve.
The broader market signal is that model routing is becoming workflow routing. As the number of capable models grows, the hard question is not always which model is strongest. It is how to choose a path based on cost, speed, reliability, and risk. Cascade's quality gate and Critique's independent review put testing, review, and human intervention inside the agent workflow rather than leaving quality to the final conversation alone.
Multi-model orchestration also creates governance work. Teams need to know which model ran at each step, who made an accept-or-escalate decision, what the critic saw, and whether a failed task can be replayed or rolled back. Because HydraFusion is still a preview, the safer way to use it is as an observable experiment with version control, tests, and permissions—not as a black box that automatically guarantees code quality.
