Meta SAM 3D is a generative neural network that reconstructs complete 3D objects and scenes from a single image, even in cluttered or occluded environments. It includes two models: SAM 3D Objects for robust object and scene reconstruction, and SAM 3D Body for accurate full human body pose, shape, and mesh estimation using the Meta Momentum Human Rig (MHR). The models support promptable inputs like masks and keypoints, enable cross-model integration for unified 3D scenes, and deliver state-of-the-art performance for AR, robotics, gaming, and creative applications.