GitHub launches Copilot Fusion, a new model selection and orchestration system that beats Claude Code for half the API cost

GitHub launched GitHub Co-pilot fusiona new system that handles model selection and orchestration for coding tasks so developers don’t have to.

By combining HydraFusion’s advanced project execution model orchestration with Copilot automatic model selection, Copilot Fusion can now manage both model selection and orchestration. GitHub reports that Fusion Max delivered comparable benchmark performance to Claude Code with Opus 5.5 Max at about half the average API cost and in about 40% less time per task.

Available now, GitHub Copilot Fusion is part of GitHub’s new vision to make GitHub “the world’s most comprehensive software delivery system,” with developers delegating model choices while retaining responsibility for the direction and results of the work.

GitHub announced the news Wednesday during GitHub Universe in San Francisco, the company’s flagship event for developers.

GitHub Copilot Fusion is now the decision maker

For tasks involving reasoning, code generation, debugging, or using tools, GitHub says developers can delegate decisions about which models to use and how they work together. His instructions: “Choose Copilot Fusion and let Copilot intelligently route and orchestrate the models it uses for you. »

This looks like a time saver, since different tasks benefit from different models based on capabilities and cost. With Copilot Fusion functioning as a “model chooser,” developers don’t have to weigh performance and price for each task.

More compelling is the idea that developers can use Copilot Fusion to strategically switch models mid-workflow, allowing the system to switch between models based on what’s best for each step.

After analyzing the needs of the task and choosing a model, the system orchestrates the next steps based on what makes the most sense. This could be a single model that does everything, a “waterfall” workflow in which a QA evaluates one model’s response and then decides whether to accept or escalate it, or a “critical” workflow in which one model reviews the work of another model.

This could reduce the mental load on developers who might inadvertently choose inefficient models when their attention drifts.

As Hari Pokhrel, principal software engineer at AT&T, describes in a statement for GitHub: “Previously, developers often chose a more powerful model for a complex task and forgot to go back to day-to-day activities. »

So what can developers decide for themselves?

Above all, the level of effort.

While Copilot Fusion handles model selection and orchestration, developers choose the level of effort: Fusion Standard for daily coding; Fusion High for more difficult tasks and deeper reasoning; Fusion Max for the most demanding and complex jobs.

That’s if you’re free and comfortable letting GitHub’s systems take the lead. If you prefer, developers can implement other controls.

According to GitHub, enterprise admins can control which templates Copilot Fusion can choose from, disabling certain templates for Copilot Fusion without disabling them for the entire developer team. Administrators can also adjust the policy state for each model independently and set whether or not Copilot Fusion is the default model for Copilot users.

Why let Copilot Fusion guide you

According to GitHub and Pokhrel’s statement, handing over model selection and orchestration to the new merge system is the smarter path for developers, allowing you to balance performance and cost without having to choose between models.

GitHub claims to have improved its performance since the HydraFusion project launched as a research preview in September 2026. On the Artificial Analysis Coding Agent Index, GitHub reports that Copilot Fusion Max scored 60, two points shy of Claude Code with Opus 5.5 Max. GitHub reports that Fusion Max achieved this for about half the average API cost per task, using about half as many tokens and taking about 40% less time per task.

In the statement provided to The new battery According to GitHub, Pokhrel credits Copilot Fusion Standard with improving token efficiency at AT&T, reducing unnecessary spending, and preserving capacity for higher value work.

Template routing becomes default routing

Copilot Fusion is a part of GitHub’s future vision, released alongside a slew of other products at the conference, including the GitHub Copilot app, Copilot Projects, and Copilot Automations.

By deferring template selection to its system, GitHub believes developers will have more time and brain space to focus on other work.

It seems like everyone is moving in this direction. This month, Ramp described moving model selection to Open SWE, its open source coding agent. This summer, Replit did the same, making its “smart template routing” system the default for every account.

Meredith Shubel is a technical writer covering cloud infrastructure and enterprise software. She has contributed to The New Stack since 2022, profiling startups and exploring how organizations are adopting emerging technologies. Beyond The New Stack, she writes white papers, executive signatures,…

Read more from Meredith Shubel