Embodied First Exam evaluates coding agents on 47 robots
November 8, 2026
The Embodied First Exam provides a unified framework for evaluating coding agents across 1,156 tasks and 35 suites using 47 different robots. It tests whether agents can use direct control, code, or trained policies to handle spatial puzzles and dexterous manipulation.
HOW THIS AFFECTS YOU
●
builderThis helps you measure how well your agentic code translates to real hardware.
●
researcherYou can use this to see if your coding agents actually work in physical environments.