AgiBot releases GO-1, its generalist embodied foundation model

AgiBot unveils Genie Operator-1 (GO-1), built on a vision-language-latent-action (ViLLA) architecture and trained on its AgiBot World dataset of over a million real robot trajectories.