Codebench

Study / Research

An internal Anthropic benchmark used to measure model progress on real-world R&D tasks within the company.

Mentioned in 1 video

Videos Mentioning Codebench

Opus 5.5: How Close Are We to Automated AI Research?

Opus 5.5: How Close Are We to Automated AI Research?

AI Explained

An internal Anthropic benchmark used to measure model progress on real-world R&D tasks within the company.