Perceptron releases Isaac 0.5: An embodied foundation model cutting teleoperation needs by 210X

rohanpaul_ai · x · 2026-08-27

Perceptron released Isaac 0.5, a 36B parameter dynamic MoE embodied foundation model with open weights. It combines multimodal video understanding, reasoning, and robot control, capable of outputting object locations, task progress, or actions directly. By scaling action-free video, it reduces the need for expensive teleoperation data by 210X. The model allows for fast one-demonstration adaptation and can be used either as a policy for robot arms or as a perception module feeding other planners.

Related event: Perceptron releases Isaac 0.5, a 36B open-weight embodied foundation model(10 posts)→

Original post →

More from Embodied

Embodied channel →