VLA models
OpenVLA
United States
The reference open-source VLA: beat the closed 55B RT-2-X with a 7B model and released all checkpoints and training code under MIT, catalyzing the academic and startup VLA ecosystem.
OrganizationStanford / UC Berkeley / TRI / Google DeepMind (academic consortium)
CountryUS
Release2024-06
Params (B)7
Open weightsyes
Architectureautoregressive VLA: Prismatic VLM (Llama-2-7B language backbone + fused SigLIP and DINOv2 vision encoders), actions as discrete tokens
LicenseMIT
Modalitiesvision, language, action
Embodimentsmanipulator, cross-embodiment
Training data970k robot manipulation episodes from the Open X-Embodiment dataset; trained on 64 A100 GPUs for 15 days
Sources : arXiv (2024-06-13)Hugging Face (model card) (2026-07-09)OpenVLA project (2026-07-09)
← Back to comparator · VLA models