π0
In plain wordsA general-purpose AI model for controlling robots, released in 2024 by the start-up Physical Intelligence, that sees through cameras, follows typed or spoken instructions and moves robot arms deftly enough to fold laundry.
A vision-language-action modelvision-language-action model, n.: An AI model that looks through a robot's cameras, reads an instruction such as 'put the cup in the sink', and turns both straight into the robot's movements. and robot foundation modelrobot foundation model, n.: One large AI model trained on huge amounts of robot and internet data so that it can control many robots on many tasks, and be adapted quickly to new ones. from the company Physical Intelligence, intended as a single generalist policy for many robot bodies and tasks. It couples a pretrained vision-language modelvision-language model, n.: An AI model that understands images and text together — it can describe a photo, answer questions about a chart or find the picture that matches a sentence. (the 3-billion-parameter PaliGemma) with a separate, roughly 300-million-parameter 'action expert' that produces continuous motor commands by flow matchingflow matching, n.: A way to train generators that learns, for every point between random noise and real data, which direction to move, so that noise can be steered smoothly into new examples. rather than as discrete tokens, emitting chunks of 50 future actions at once so that dexterous tasks can be controlled at up to 50 Hz (action chunkingaction chunking, n.: Having a robot's AI plan a short burst of upcoming movements, say the next second, all at once instead of one tiny motion at a time, which makes motion smoother and mistakes rarer.). Pretraining combined more than 10,000 hours of the company's own teleoperated data, covering 7 robot configurations and 68 tasks, with the Open X-EmbodimentOpen X-Embodiment, prop. n.: A large shared collection of robot training data, pooled by dozens of research labs in 2023 and covering many kinds of robots, so that one AI model could learn from all of them. collection; post-training on curated data taught long multi-stage tasks such as putting together cardboard boxes, clearing dishes from a table and unloading a dryer to fold the clothes. Its weights and code were released openly in early 2025 (openpi), making it a common baseline in robot-learning research.
We fine-tuned π0 on twenty hours of our own teleoperation data and it learned to pack the parts trays.
Origin
Announced on 31 October 2024 by Physical Intelligence, a San Francisco robotics start-up founded in 2024 by a group including Karol Hausman, Sergey Levine and Chelsea Finn, in the paper 'π0: A Vision-Language-Action Flow Model for General Robot Control' (Kevin Black and colleagues). Successors include π0-FAST and π0.5 (2025), the reinforcement-learning-refined π*0.6 (late 2025) and π0.7 (16 April 2026).
Usage note
Usually written π0 or pi0. Its demonstrations featured unusually long, dexterous task sequences, but generalization to unseen homes, objects and robot bodies remained limited and became the explicit target of later versions.
Related
- vision-language-action modelvision-language-action model, n.: An AI model that looks through a robot's cameras, reads an instruction such as 'put the cup in the sink', and turns both straight into the robot's movements.
- robot foundation modelrobot foundation model, n.: One large AI model trained on huge amounts of robot and internet data so that it can control many robots on many tasks, and be adapted quickly to new ones.
- flow matchingflow matching, n.: A way to train generators that learns, for every point between random noise and real data, which direction to move, so that noise can be steered smoothly into new examples.
- action chunkingaction chunking, n.: Having a robot's AI plan a short burst of upcoming movements, say the next second, all at once instead of one tiny motion at a time, which makes motion smoother and mistakes rarer.
- generalist robot policygeneralist robot policy, n.: A single robot-control program trained to do many different tasks in many places, often on several kinds of robots: the opposite of a specialist trained for one job.
- Open X-EmbodimentOpen X-Embodiment, prop. n.: A large shared collection of robot training data, pooled by dozens of research labs in 2023 and covering many kinds of robots, so that one AI model could learn from all of them.
- Gemini RoboticsGemini Robotics, prop. n.: A family of AI models from Google DeepMind, built on its Gemini models, that let robots understand what they see and what people ask of them, and turn that understanding into physical actions.
- teleoperationteleoperation, n.: Controlling a robot or machine from a distance, with a person steering it in real time using a joystick, a matching 'puppet' arm, a VR headset or a screen.
- Gemini RoboticsGemini Robotics, prop. n.: A family of AI models from Google DeepMind, built on its Gemini models, that let robots understand what they see and what people ask of them, and turn that understanding into physical actions.
Sources
- Black, K., Brown, N., Driess, D., Esmail, A., Equi, M. et al. (2024). π0: A Vision-Language-Action Flow Model for General Robot Control. arXiv:2410.24164. arxiv.org/abs/2410.24164
Mentioned in
Chelsea FinnChelsea Finn, prop. n.: An American computer scientist at Stanford known for 'learning to learn' methods that let AI adapt to new tasks from a few examples, and a co-founder of the robotics company Physical Intelligence., Gemini RoboticsGemini Robotics, prop. n.: A family of AI models from Google DeepMind, built on its Gemini models, that let robots understand what they see and what people ask of them, and turn that understanding into physical actions., Physical IntelligencePhysical Intelligence, prop. n.: A San Francisco company, founded in 2024, that builds general-purpose AI 'brains' meant to control many different kinds of robots doing many different tasks., Sergey LevineSergey Levine, prop. n.: A Berkeley computer scientist, one of the most prolific researchers on how robots can learn skills from data and experience, and a co-founder of the robotics company Physical Intelligence.