The brief
- Sets the research pace: π0 (2024), π0.5 (2025), π*0.6 (2025) and π0.7 (Apr 2026).
- Valued at about $11B in 2026, roughly double 4 months earlier, with no revenue.
- Its open weights (openpi) are the default starting point for many robot labs.
Technical approach
Image-and-language AI plus an action module
A 3B-parameter vision-language model is paired with a 300M-parameter action module that outputs smooth motion up to 50 times a second.
Trained on many robot bodies
Trained on thousands of hours of remote-controlled data from many robot types plus web images and text, so one model drives different robots.
Works in unseen homes
π0.5 mixes robot data, web data and step-by-step instructions, and cleaned kitchens and bedrooms in homes it had never seen.
Learning from experience
π*0.6 learns from its own attempts and human corrections, not just copying. It roughly doubled speed on tasks like making espresso.
Physical Intelligence, one program
Updated 16 Apr 2026
Research & demos
- 16 Apr 2026paperπ0.7: a steerable generalist robotic foundation model with emergent capabilities (opens arxiv.org)π model family
Early signs of combining skills, like folding laundry on a robot never trained on laundry.
- Nov 2025
Learning from real deployment data made it faster and more robust than copying demos alone.
- 22 Apr 2025
Showed mobile robots cleaning homes that were not in the training data.
- Feb 2025
Made a top robot model free for anyone to use and build on.
- Jan 2025
A new way to compress robot motions into tokens that made some robot models train much faster.
- 31 Oct 2024paperπ0: a vision-language-action flow model for general robot control (opens arxiv.org)π model family
Set the recipe: an image-and-language AI plus an action module, for skilled control of many robot types.
Open technical problems
- Reliability for paying customersActive research
Success in new places is impressive for research but far below what customers need. Closing the gap likely needs learning on real jobs.
- Data scalingOpen
Robot data is far scarcer than text. Whether remote control, human video or simulation scales best is the big open question.
Peers
Company & funding
- Valuation
- $11Bprivate
- Raised
- $2.1Bdisclosed private rounds
- Last round
- Growth round2026
- HQ
- US
- Founded
- 2024
Post-money valuation, log scale
| Date | Round | Amount | Post-money | Lead |
|---|---|---|---|---|
| 2026 | Growth round | $1B | $11B | Founders Fund |
| Nov 2025 | Series B | $600M | $5.6B | CapitalG |
| Nov 2024 | Series A | $400M | $2.4B | — |
| Mar 2024 | Seed | $70M | — | Thrive Capital |
Partners & customers
- Investor
- Investor
- Investor
Led the 2026 round at about $11B.
- Investor
Led the Nov 2025 $600M round.
People
- Karol Hausman (opens karolhausman.github.io)Co-founder & CEO
- Sergey Levine (opens en.wikipedia.org)Co-founder
- Chelsea Finn (opens en.wikipedia.org)Co-founder