Starting today, the new Claude 3.5 Sonnet begins rolling out on GitHub Copilot, allowing developers to choose Claude 3.5 Sonnet for coding – directly in Visual Studio Code and GitHub.com. This integration brings Claude's coding capabilities to the GitHub community of over 100 million developers.

The updated Claude 3.5 Sonnet outperforms all publicly available models in SWE-Bench Verified, a benchmark that measures how models solve real-world GitHub problems. It also achieves the highest score in its class (93.7%) on HumanEval, a benchmark that tests the ability to write correct Python functions from natural language descriptions.

Claude 3.5 Sonnet will be available in public preview to all GitHub Copilot Chat users and organizations in the coming weeks.

Use cases

With context to your entire codebase, you can use Claude 3.5 Sonnet on GitHub Copilot to:

• Write production-ready code: Ask GitHub Copilot Chat to convert natural language descriptions into high-quality code that conforms to your project's patterns and conventions.
• Debugging with inline chat: Highlight problematic code for immediate analysis and fixes, or ask GitHub Copilot Chat to explain error messages and suggest solutions.
• Create tests from implementation: Select any function to automatically generate comprehensive test suites that fit your testing framework and cover edge cases.
• Understanding code with contextual explanations: Hover over functions or highlight blocks of code in VS Code to get explanations of what the code does and how it fits into your project.

First steps

GitHub is releasing Claude 3.5 Sonnet in public preview starting today. Claude 3.5 Sonnet runs on GitHub Copilot over Amazon Bedrock and leverages Bedrock's cross-region inference to further improve reliability.

All developers and organizations should have access to Claude 3.5 Sonnet in the coming weeks. Explore the GitHub blog and documentation to learn more.