A vision-language-action model (VLA) is a robot policy that takes camera images and a natural-language instruction, usually with proprioceptive state, and outputs robot actions. It is built by adapting a vision-language model (VLM) pretrained on Internet-scale data. The term comes from the RT-2 paper (Brohan et al., 2023): "We refer to such category of models as vision-language-action models (VLA)". RT-2 co-fine-tuned a VLM on robot trajectories and web data and emitted actions as text tokens.
VLAs differ mainly in how they produce actions. Autoregressive VLAs such as OpenVLA and π0-FAST predict discrete action tokens. Others attach a continuous action head to the VLM: π0, π0.5, SmolVLA, GR00T N1.7, X-VLA, EO-1 and EVO-1 use flow matching, while OpenVLA-OFT uses parallel decoding with L1 regression. Size varies widely, from OpenVLA's 7B parameters to SmolVLA, which is "designed to be trained on a single GPU and deployed on consumer-grade GPUs or even CPUs".
Two misconceptions recur. "VLA" names an architecture family, not a capability level. Language conditioning does not mean the model follows arbitrary instructions on a new robot, and deployment on a robot absent from pretraining generally requires fine-tuning (see embodiment gap). Nor does the label imply a size: parameter count, action head and precision, not the acronym, decide whether a VLA fits on a Jetson Orin Nano or needs an H100.