π0.7Physical Intelligence
Robot foundation models
Physical Intelligence · AI model
π0.7
April 2026 'steerable' generalist model. It accepts several kinds of prompts (language, speed and quality hints, control-mode labels and picture subgoals) and is reported to combine skills it already knows to do tasks it was never trained on, such as folding shirts on a robot arm that had no shirt-folding data.
- Kind
- Robot foundation models
- Maker
- Physical IntelligenceUnited States
- First year
- 2026
- Weights
- —Not published
- Robots
- 1Named as using it
- Country
- United States
Robots that use it
1 robot model named by the maker or reputable press.
Not yet in the atlas
- Universal Robots UR5e (shirt-folding example reported in coverage)forklog.com
Other robot foundation models
Newest first.
- 2026Isaac GR00T N1.7NVIDIA · 2 robot models
- 2026Helix 02 (and Helix 2.5)Figure AI · 1 robot model
- 2026Gemini Robotics 2 (with ER 2 and On-Device 2)Google DeepMind · 1 robot model
- 2025Isaac GR00T N1NVIDIA · 1 robot model
- 2025Isaac GR00T N1.5NVIDIA · 1 robot model
- 2025Isaac GR00T N1.6NVIDIA
- 2025π0.5Physical Intelligence · 1 robot model
- 2025π*0.6 (pi-star 0.6)Physical Intelligence
- 2025HelixFigure AI · 1 robot model
- 2025Gemini Robotics and Gemini Robotics-ERGoogle DeepMind · 3 robot models
More from Physical Intelligence
Sources
From the maker, its papers and reputable press.
Statistics only, not advice. No affiliation, sponsorship or endorsement. How we work · Terms