GitHub Copilot Fusion orchestrates between models at half the cost (twitter.com)

GitHub has launched Copilot Fusion within GitHub Copilot, a new runtime orchestration system that adeptly coordinates between AI models from Anthropic and OpenAI to optimize performance, cost, and latency for various coding tasks. In a competitive benchmark, Copilot Fusion achieved a score just 2 points below Anthropic's Opus 5.5, yet does so at a significantly lower cost of $5.59 per task, compared to $11.42 for Opus 5.5 in Claude Code. This innovation positions GitHub as a frontrunner in the AI coding assistant landscape, providing developers with a cost-effective solution that maintains competitive performance. The implications of Copilot Fusion are noteworthy for the AI/ML community; it enables more efficient use of resources by leveraging multiple models, ultimately reducing the average token usage for tasks—about half compared to Claude Opus 5.5 and 30% less than another competitive system. Additionally, with leading scores in specific benchmark categories like Terminal-Bench 4.0, this release not only enhances developer productivity but also accelerates the evolution of AI-driven coding tools, highlighting a trend toward increasingly optimized and accessible AI solutions in software development.