GitHub Enhances Copilot with Custom Model for Improved Code Completions
GitHub has rolled out a significant upgrade to its AI-powered coding assistant, Copilot, introducing a custom model designed to deliver faster and more accurate code completions. The enhancements, driven by extensive developer feedback, focus on real-world usability and efficiency.
The updated model boasts a 20% increase in accepted and retained characters, a 12% higher acceptance rate, and a threefold improvement in token-per-second throughput. Latency has been reduced by 35%, streamlining the developer experience across various coding environments.
This shift from mere acceptance rates to optimizing for code FLOW and retained characters reflects GitHub's commitment to productivity. The improvements ensure more of Copilot's suggestions make it into final code, reducing editing time and allowing developers to focus on building.