Autoregressive π₀ policy using the FAST action tokenizer.
Research, evaluation and robot-specific training of π₀-FAST tokenized-action policies.
A newer version is available: π₀.₇ ↗
Access and availability
- Code
- Public access
- Model files
- Public access
Training / fine-tuning code
- Access
- Public access
Verified on · 2026-10-30
- Code license
- Apache-2.0
Model files
| File | Access | Weights license |
|---|---|---|
| lerobot/pi0fast-base | Access |
Getting started
- Runtime and frameworks
- openpi / LeRobotNVIDIA GPU · bfloat16
- 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 model
- Backbone
- PaliGemma vision-language backbone with FAST autoregressive action tokenizer
- Weight parameters
- 2.9233 B
- Input
- Language instruction / RGB camera images / Robot state / proprioception
- Output
- Tokenized robot action chunks
- Released
- January 16, 2025
Related robots
Other versions in this model family
Family and other-version demonstrations are context, not independently verified results for the selected weights.
Technical details
Tokenized robot action chunks
Languages
English
Functional purpose
Action & control
Pipeline task
Robotics
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.
Submit a source or correction
