Physical Intelligence
Robotics companies and developers seeking general-purpose AI models to control robots for a wide range of tasks.
Overview
Physical Intelligence (π) is a San Francisco-based AI company founded in 2024 that develops general-purpose AI models for robots. The company brings together engineers, scientists, roboticists, and company builders to create learning algorithms that aim to control any robot to perform any task. It has released a series of robotic foundation models, including π0, π0.5, π0.6, and π* 0.6, and has open-sourced the weights and code for π0. The company is backed by investors including Bond, Jeff Bezos, Khosla Ventures, Lux Capital, OpenAI, Redpoint Ventures, Sequoia Capital, CapitalG, and Thrive Capital.
Funding: Total funding of $1.07B over 3 rounds, with the first round on Mar 12, 2024 (per Tracxn).
What it does
Physical Intelligence builds vision-language-action (VLA) models that enable robots to perceive, reason, and act in the physical world. Its models, such as π0 and π0.5, are trained on large-scale multi-task and multi-robot data and can control mobile manipulators to perform tasks like cleaning a kitchen or bedroom. The company also develops methods for efficient robot action tokenization, real-time action chunking, and reinforcement learning to improve model performance and generalization.
Key features
- Generalist robot policies (π0, π0.5, π0.6, π* 0.6) with open-world generalization
- Vision-language-action (VLA) models that combine internet-scale knowledge with robotic control
- Multi-scale embodied memory (MEM) for long and short-term memory in robots
- Reinforcement learning methods for precise manipulation and improved throughput
- Real-time action chunking for large VLAs with high latency
- Open-sourced weights and code for π0 and π0-FAST
- Human-in-the-loop feedback for step-by-step task reasoning
Use cases
- Mobile manipulation tasks such as cleaning a new kitchen or bedroom
- Precise manipulation tasks requiring fast online reinforcement learning
- Complex tasks lasting longer than ten minutes using long-term memory
- Difficult manipulation challenge tasks solved via fine-tuning
- Transfer of human video demonstrations to robotic tasks
Pricing
No pricing information is publicly available on the company's website or in search results.
Pricing is gathered from the company's public pages and may change. Check the vendor's site before buying.
Pros and cons
Strengths
- Backed by prominent investors including OpenAI, Sequoia Capital, and Jeff Bezos
- Open-sourced π0, enabling community access and development
- Demonstrated step-change in generalization with π0.7
- Strong research output with regular model releases and technical blog posts
- Focus on general-purpose AI for physical world, not narrow single-task robots
Limitations
- No pricing information publicly available; likely enterprise-focused with contact sales
- No self-serve or free tier for developers
- Early-stage company (founded 2024) with evolving product line
- Models may require significant technical expertise to deploy
What sets it apart
- General-purpose robotic foundation models designed to control any robot for any task
- Open-sourcing of π0 weights and code, a rare move in the robotics AI space
- Strong backing from a wide range of top-tier investors including OpenAI and Jeff Bezos
Frequently asked questions
What is Physical Intelligence?
Physical Intelligence is a San Francisco-based AI company founded in 2024 that develops general-purpose AI models for robots, aiming to control any robot to do any task.
What are the company's main products?
The company has released several robotic foundation models, including π0, π0.5, π0.6, and π* 0.6, which are vision-language-action models for robot control.
Is Physical Intelligence's technology open source?
Yes, the company has open-sourced the weights and code for π0 and its π0-FAST autoregressive model.
Who are the investors in Physical Intelligence?
Investors include Bond, Jeff Bezos, Khosla Ventures, Lux Capital, OpenAI, Redpoint Ventures, Sequoia Capital, CapitalG, and Thrive Capital.
What can the robots do with Physical Intelligence's models?
The models can control mobile manipulators to perform tasks like cleaning a new kitchen or bedroom, and can handle precise manipulation tasks with reinforcement learning.
Sources
This profile was compiled from the company's own pages and public web research.
Last researched August 16, 2026.
Manage this listing
Own this company? Update your listing information.