OpenBMB's first embodied series (Apache 2.0): RobotManip, a 1.5B generalist VLA using MiniCPM-V visual-token compression (60-frame streaming at 120ms/step, claimed to beat π0.5-3B), and RobotTrack, a 0.4B person-following VLA running 5+ FPS on a Unitree Go2. No paper yet.

Model Details

Architecture DENSE

Variants

Name Parameters Notes
MiniCPM-RobotManip 1.5B
MiniCPM-RobotTrack 0.4B
on-deviceopen-weightrobotics

Related