claw-bench

Measure AI agents’ performance with standardized tests across 314 tasks, 33 domains, and 4 difficulty levels for clear, reproducible comparison.