opentau.policies.pi06

PI06 Policy Module.

This module implements the π06 (Pi06) Vision-Language-Action model from Physical Intelligence. Relative to π05, π06 swaps the PaliGemma-3B vision-language backbone for a Gemma-3 4B multimodal backbone, enlarges the action expert to ~860M parameters (matching the backbone depth), raises the default image resolution from 224×224 to 448×448, and cuts the flow-matching denoising schedule from 10 to 5 steps.

References

  • π0.6 Model Card, Physical Intelligence, November 17, 2025.

  • π*0.6: a VLA That Learns From Experience, arXiv:2511.14759.

Modules

configuration_pi06

Configuration module for the PI06 Policy.

gemma3_with_expert

Gemma 3 backbone with Gemma-v1 action expert, for the PI06 policy.

modeling_pi06

π06: a Vision-Language-Action model built on Gemma 3 4B.