How does Copilot Fusion choose which AI model to use?
GitHub launched Copilot Fusion at GitHub Universe, its developer conference, on Wednesday. The system decides which AI model, or mix of models, should handle each step of a coding task. It weighs what the task needs across reasoning, code generation, debugging and tool use.
Jay Parikh, executive vice president of CoreAI at Microsoft, described the system in a GitHub blog post. He wrote that Fusion Max scored within two points of Anthropic's Claude Opus 5.5 Max on the Artificial Analysis Coding Agent Index. It did so at half the cost and with 40% less time per task. In a LinkedIn post, Microsoft chief executive Satya Nadella said Fusion delivers frontier-level quality 40% faster and at up to 77% lower cost.
Nadella also said the unified system that runs Copilot's agents is now available in Foundry, Microsoft's managed service for building long-running agents. The launch comes a few months after a group of coders lost their DMCA case against Copilot at the Ninth Circuit in September.
Copilot without a separate subscription
Does Copilot Fusion cost extra for GitHub Enterprise customers?
From Wednesday, GitHub Enterprise customers can use every Copilot feature without buying a separate subscription. They pay for the AI credits they use. The offer currently runs through GitHub's sales team, with self-serve access coming soon. Existing Copilot Business and Copilot Enterprise customers can keep their plans or switch when they choose.
GitHub also showed the GitHub Copilot app, which connects agents and tools across projects. Inside the app, Copilot projects run agent work across several repositories in parallel managed sessions. Session continuity lets a single cloud agent session move between the app, VS Code, the command line and the web.
Copilot automations can now learn from their results and propose changes for developers to approve. Microsoft added coding to its own Copilot app, a separate product, in September.
For code review, GitHub added Copilot code review Max. The company says it posted the highest F1 score on ReviewBench for high-severity issues. The category is crowded. Startups are selling AI code review too, and CodeRabbit raised $143 million in August to check code written by AI.
More capacity for agents
This year GitHub has added 4 million CPU cores and 160 petabytes of high-speed storage, Parikh wrote. It has also added 50 terabits per second of network capacity. GitHub is moving to a multi-region deployment on Microsoft's Azure and expects to finish the work early next year. Nadella said the change will make GitHub more resilient.
Parikh also made a prediction about the scale of what is coming. “I think all the software we’ve written to date will amount to less than 1% of the software we’ll write in the next five years,” he wrote.
What the Octoverse numbers show
GitHub published its annual Octoverse report the same day. Developers pushed code 17.9 billion times in the 12 months ending September 30, more than in 2024 and 2025 combined. Commits reached 34.3 billion, and merged pull requests reached 1.3 billion.
How many pull requests do AI agents start?
By September, more than 40% of pull requests started with a draft from an agent, up from 17% in March. That amounted to 110 million pull requests in September alone. Activity rose sharply around June, when agents began working across days and running in the cloud, according to the report. GitHub says its agent count is a floor because it cannot detect every agent.
The platform now has more than 225 million developers. The United States accounts for 38.3 million of them and India for 32 million. In September, 61% of Copilot users who chose their own model used more than one. TypeScript remained the most used language. Rust grew fastest among the top languages, doubling in a year.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.