ModelFeatured
π0 (pi0)
Physical Intelligence generalist robot foundation model with flow-matching action generation.
ModelFeaturedFully open
Octo
Generalist robot policy Transformer with multi-modal inputs and multi-embodiment finetuning.
Why: Open generalist policy that is friendly to finetune
Overview
Octo is a generalist robot policy transformer trained across diverse robot datasets (OXE-scale). It is a standard open baseline for cross-embodiment finetuning and language-conditioned control experiments.
Read the model card for observation spaces and finetune scripts. For many labs it is the “open generalist” checkpoint before proprietary VLAs.
Learning Path: core VLA-stage generalist baseline.
Who it is for
Labs needing lighter generalist policies; multi-robot finetune studies.
Key highlights
- Compact generalist policy
- Multi-embodiment finetune story
- Open weights and code
When to use
- Limited GPUs
- Cross-robot finetuning research
- Baselines vs larger VLAs
When not to use
- You specifically need a large VLM backbone for language reasoning
Getting started
- 1Load Octo weights
- 2Run finetune on a single-embodiment demo set
- 3Compare to from-scratch BC
Papers & reading
How it compares
Caveats & pitfalls
- Smaller capacity can underfit highly diverse language instructions.
Content reviewed 2026-07-28