According to @DeepLearningAI, Meta released DINOv3, a self-supervised vision transformer that improves image embeddings for tasks like segmentation and depth estimation (source: DeepLearning.AI). The model has 6.7 billion parameters and was trained on over 1.7 billion Instagram images, highlighting a significant scale-up in self-supervised vision pretraining (source: DeepLearning.AI). Technical updates include a new loss term that preserves patch-level diversity, mitigating limitations from training without labels and strengthening downstream performance baselines (source: DeepLearning.AI). Weights and training code are available under a license that allows commercial use but forbids military applications, enabling broad enterprise deployment while constraining defense use cases (source: DeepLearning.AI). The source does not cite any direct cryptocurrency market impact; traders can note that a stronger open self-supervised backbone may influence developer adoption trends in AI infrastructure that markets often track for sentiment, but no market effects are stated by the source (source: DeepLearning.AI).