23 Daily — According to thebridge.jp, on October 28, 2026, GitHub announced a new feature within its AI coding assistant GitHub Copilot called "Copilot Fusion." This feature automatically selects and combines multiple AI models according to different coding tasks, enhancing coding support accuracy and efficiency. Copilot Fusion is now generally available to all Copilot users.

Copilot Fusion Automates Multi-Model Selection Based on Tasks, Supporting Various Scenarios

Copilot Fusion is accessible via Visual Studio Code, the GitHub Copilot app, GitHub Copilot CLI, and JetBrains IDEs, covering all Copilot plans. Instead of manually selecting a single model, users choose one of three tiers according to the complexity of their tasks: Fusion Standard for everyday coding, Fusion High for difficult tasks requiring deep reasoning, and Fusion Max for the most demanding workloads. The feature integrates the previously released Project HydraFusion multi-model collaboration and Copilot's automatic model selection, deciding the optimal models and their usage for each code prompt.

Three Model Operation Modes Optimize Performance, Significantly Reducing Costs and Time

Copilot Fusion operates in three modes: Single, where one model responds directly; Cascade, where an efficient model answers first and, if the answer is insufficient, it is escalated to a stronger model; and Critique, where models review and correct each other's responses. According to the Coding Agent Index by Artificial Analysis, Fusion Max scores 60 points, within two points of Claude Code using Opus 5.5 Max. Moreover, average API costs and token usage per task are about half, with time reduced by roughly 40%.

Enterprise Policy-Driven Model Management Cuts Waste and Costs

Copilot Fusion only uses models enabled by enterprise or organizational policy. Administrators can exclude certain models from Fusion's candidate list without blocking the models entirely, and specify the default Fusion tier for Copilot using Enterprise managed settings. Hari Pokhrel from AT&T noted that previously developers often forgot to switch back from powerful models used for complex tasks to normal models for everyday work, leading to wasteful expenses; with Fusion Standard automatically selecting the appropriate model by task, such unnecessary costs are reduced.

Editor's take: Copilot Fusion offers a powerful efficiency tool for startup teams, with verification needed through hands-on use

GitHub Copilot Fusion's task-oriented automatic selection and combination of models marks a significant advance in improving AI coding assistance accuracy and efficiency, accommodating varied work demands. For startups, this reduces the complexity and time costs of adjusting models manually, boosting development productivity. Entrepreneurs can validate benefits by directly using Copilot Fusion, assessing code quality and speed improvements, and measuring cost-effectiveness in their development workflows. This practical verification is critical for adopting multi-model AI tools.