Robot AI model
π0.7
π0.7 is a generalist VLA steered by language, episode metadata, control mode and optional subgoal images, with a video-history encoder. The paper reports approximately 5B total parameters and evaluations on mobile, static bimanual and UR5e platforms.
Records last verified Oct 13, 2026 · Send a correction
Also known as pi0.7
- Organization
- Physical Intelligence
- Type
- VLA
- Released / introduced
- Apr 16, 2026
- Public weights
- Not established
- Licence / scope
- Not established
- Parameters
- 5B
- Base model
- Gemma 3 4B
- Humanoid association
- None reported
- Inputs
- vision, language, proprioception, subgoal images, episode metadata, control mode
- Outputs
- continuous robot action chunks
Release dates can mark an announcement or paper introduction. Public checkpoints, code licences and weight terms are distinct; follow the publisher’s links for the stated scope.
Documented embodiments
Source-associated hardware; this list does not establish plug-and-play compatibility or commercial robot capability.
- Bimanual mobile manipulator
Two 6-DoF arms; exact platform model is not identified.
- BiPi
Static bimanual system with lightweight 6-DoF arms.
- Bimanual UR5e
Robotiq parallel-jaw grippers; cross-embodiment evaluation.
- Single-arm BiPi-derived setup
Used in generalization and language-following experiments.
Disclosed training data
- Multi-robot demonstrations, autonomous evaluations, human interventions, open robot datasets, egocentric human videos and web data
More robot AI models
Evidence
Sources
- [s1] π0.7: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities ↗
Physical Intelligence · paper · published Apr 16, 2026 · accessed Oct 13, 2026
- [s2] π0.7 paper: architecture, parameter scope, training and embodiments ↗
Physical Intelligence · paper · published Apr 24, 2026 · accessed Oct 13, 2026