DINOv3 scales self-supervised learning (SSL) for images to produce our strongest universal vision backbones, enabling breakthrough performance across...