Mercor releases OWE-Bench engineering benchmark

Mercor has released OpenWorldEngineer (OWE-Bench), a frontier coding benchmark featuring 26 long-horizon engineering and research tasks across software engineering, physical engineering, machine learning, and scientific computing. Graded on continuous scales using unseen held-out cases, the benchmark tests whether AI agents can complete complex projects from scratch or codebases. Claude Fable 5.2 currently leads the leaderboard with a score of 54.6%, followed by GPT-6 Astra and Claude Opus 5.5. Open article →