GitHub Copilot optimizes VS Code token efficiency to reduce costs and latency (opens in new tab)

GitHub has implemented several optimizations in VS Code to improve the token efficiency of agentic Copilot sessions. These enhancements address the move toward usage-based billing by reducing the volume of data sent to models and maximizing the use of cached states. The primary focus is on minimizing the "prompt prefix" overhead, which includes system instructions and repository context that often repeat across multiple turns. <a href="