RT-2-X is a member of the RT-X model family, trained on the collaborative Open X-Embodiment Dataset. It extends the original RT-2 VLA model by fine-tuning on cross-embodiment data from 34 institutions, enabling generalization across diverse robot platforms.