Apple paper reports zero-shot robot manipulation results with AMP1
Overview
Apple researchers report that AMP1, a harness that lets a general AI model, Claude Opus 5.5, control a robot arm directly, succeeded in 85.6% of RoboLab-120 episodes without demonstrations or task training.
The best vision-language-action model reached 49.2% and the best world-action model 42.9%, according to the paper as reported on X by Rohan Paul.
On a real Franka arm, AMP1 succeeded in 104 of 108 trials. The robot waits for the model for much of each episode, according to the same report.
Written by AI from the articles below · updated Nov 9, 9:23 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
Rohan Paul@rohanpaul_aiXPickApple paper finds a general AI model controlling a robot arm beats specialist policiesAIApple researchers report that AMP1, a harness letting a general AI model like Claude Opus 5.5 control a robot arm directly, succeeded in 85.6% of RoboLab-120 episodes zero-shot. That compares with 49.2% for the best vision-language-action model and 42.9% for the best world-action model. On a real Franka arm, the system succeeded in 104 of 108 trials, though the robot waits much of each episode for the model.

Heat Trend
Not enough continuous observations to show a trend yet.