OpenVLA is a 700-million-parameter open-source VLA model pre-trained on 970k robot episodes from the Open X-Embodiment dataset. This model sets a new industry standard for generic robot operation policies, enabling out-of-the-box control of multiple robots and rapid adaptation to new robot setups through parameter-efficient fine-tuning. OpenVLA's checkpoints and PyTorch training procedures are completely open-source, allowing the model to be downloaded and fine-tuned from HuggingFace.