BOT-AGI

The ultimate intelligence benchmark is not games.
It's physical ability.

About

How good are the leading models at controlling robots?

BOT-AGI-1 is an independent robotics benchmark, with tasks that humans can easily solve.

Tasks
Unitree G1 cube task simulation

Full benchmark coming soon

Leaderboard
Coming soon

View Qwen 3.5VL 235B replay →

Contribute

Interested in contributing tasks, evaluations, or model results to BOT-AGI-1?

Get in touch