π0.6 (pi-zero-6) is a vision-language-action (VLA) robotic foundation model developed by Physical Intelligence. It is the next version in the π line — a refinement of the π0.5 model on a slightly larger backbone — and serves as the decision layer (brain) of robots, including Ultra Robotics' stationary OP1 humanoid.
The π*0.6 (pi-star-0.6) variant is trained with the Recap method ("RL with Experience & Corrections via Advantage-conditioned Policies") — a three-stage process combining learning from demonstrations, expert corrections and autonomous reinforcement learning from the robot's own experience. Training on autonomous experience more than doubles throughput on the hardest tasks and cuts failure rates by at least 2x.
The model was demonstrated on complex, multi-hour real-world tasks: making coffee/espresso (running from 5:30am to 11:30pm), folding laundry (50 novel items in a new environment) and assembling 59 boxes on a real chocolate-packaging line, achieving over 90% success rates. The work was published on 17 November 2025.
An API Library is a software package that exposes programmatic interfaces for communicating with a device, service, or system. In robotics it typically forms a lightweight integration layer built on top of the manufacturer's official API or an open-source project, abstracting low-level protocol details and providing language-native bindings (Python, C++, Java, etc.).
The robot learning role describes software for training a robot's control policies and manipulation/locomotion skills using machine learning methods. It covers: reinforcement learning in simulation with sim-to-real transfer, imitation learning and learning from demonstration, training Vision-Language-Action (VLA) models, and fine-tuning robotics foundation models. It typically uses massively parallelized simulation environments (Isaac Lab, MuJoCo) to generate training data, then deploys the trained policies on a physical robot.
Robot Control denotes the role of software responsible for motion control, command execution, coordination of actuating elements and the direct operational logic of the robot.
Decision layer of the OP1 humanoid (Ultra Robotics). π*0.6 demonstrations: making espresso (5:30am–11:30pm), folding laundry (50 novel items), assembling 59 boxes on a chocolate-packaging line.
License family: Proprietary – Commercial
No version history.
