
Physical Intelligence’s π₀ vision-language-action base model generates continuous actions with flow matching. Pre-training combines the π dataset, covering seven robot configurations and 68 tasks, with OXE and other public data. It supports robot- and task-specific fine-tuning.
Research, evaluation and robot-specific training of π₀ vision-language-action policies.
A newer version is available: π₀.₇ ↗
Access and availability
- Code
- Public access
- Model files
- Public access
- Training data
- Partly available
Training / fine-tuning code
- Access
- Public access
Verified on · 2026-10-30
- Code license
- Apache-2.0
Training data
Some training datasets are public; this does not establish access to the complete training mixture.
Model files
| File | Access | Weights license |
|---|
Getting started
- Runtime and frameworks
- LeRobot / PyTorchCUDA-capable GPU
- Fine-tuning requirements
- LeRobot integration: use the linked policy-specific setup and training instructions; compatibility with every upstream checkpoint is not implied.
At a glance
- Model type
- Vision-language-action
- Weight parameters
- 3.5014 B
- Input
- Optional language instruction / Multi-view images / Robot state / proprioception
- Output
- Continuous actions
- Released
- October 31, 2024
Related robots
Evaluation results
No verified independent benchmark results are recorded for this version or the selected weights.
Publisher evaluation · Reported; not independently reproduced
VLABench · Primitive FT LeRobot Video ↗Overall: 44.5 score
Technical details
Architecture
Flow matching
Flow-matching policy producing continuous robot actions.
Languages
English
Functional purpose
Action & control
Training data
Model card
The paper reports 3.3B architecture parameters; the LeRobot checkpoint contains 3,501,372,176 parameters, approximately 3.50B. The model was announced on 2024-10-31; openpi code and weights were released on 2025-02-04. LeRobot provides a port and weight distribution. https://arxiv.org/html/2410.24164v1 https://www.pi.website/blog/openpi https://huggingface.co/lerobot/pi0_base
Before you use this version: Requires task- and embodiment-specific data, configuration and validation. Published examples and simulated results do not establish safe operation on a physical robot. Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.