/
#89
Physical Intelligence
Robotics foundation model lab. Builds the π (pi) model family, Vision-Language-Action models built on PaliGemma, trained on ~10,000 hours of cross-embodiment data across 7 robot platforms and 68 tasks, using flow matching for action generation. Latest model, π0.7, shows emergent generalization to tasks outside its training distribution. Code and weights are open-sourced via the openpi GitHub repo.
Links