Model Family
OpenVLA
Open-weights vision-language-action model for robot policies, 7B parameters.
Definition
OpenVLA is a 7B-parameter Llama-2-based vision-language-action model trained on the Open X-Embodiment dataset. Released openly in 2024, it lets researchers fine-tune generalist robot policies without retraining a frontier VLM from scratch.
Common use cases
- Manipulation research
- Sim-to-real
- Imitation learning