OpenVLA is a 7B-parameter open-source vision-language-action (VLA) model trained on 970k robot episodes from the Open X-Embodiment dataset. It sets a new state-of-the-art for generalist robot manipulation policies, outperforming closed models like RT-X (55B parameters) by 16.5% in task success rate across 29 tasks and multiple robot embodiments. OpenVLA supports controlling multiple robots out-of-the-box and can be quickly adapted to new robot domains via parameter-efficient fine-tuning. The model and its training pipeline are fully open-source, and models can be downloaded and fine-tuned from HuggingFace. The paper also explores efficient fine-tuning strategies, demonstrating that OpenVLA can be fine-tuned on consumer GPUs using modern low-rank adaptation methods and quantization techniques without compromising performance. The release of model checkpoints, fine-tuning notebooks, and the PyTorch codebase aims to enable future research on VLA training, data mixtures, objectives, and inference.OpenVLA is a 7B-parameter open-source vision-language-action (VLA) model trained on 970k robot episodes from the Open X-Embodiment dataset. It sets a new state-of-the-art for generalist robot manipulation policies, outperforming closed models like RT-X (55B parameters) by 16.5% in task success rate across 29 tasks and multiple robot embodiments. OpenVLA supports controlling multiple robots out-of-the-box and can be quickly adapted to new robot domains via parameter-efficient fine-tuning. The model and its training pipeline are fully open-source, and models can be downloaded and fine-tuned from HuggingFace. The paper also explores efficient fine-tuning strategies, demonstrating that OpenVLA can be fine-tuned on consumer GPUs using modern low-rank adaptation methods and quantization techniques without compromising performance. The release of model checkpoints, fine-tuning notebooks, and the PyTorch codebase aims to enable future research on VLA training, data mixtures, objectives, and inference.