multimodal AI

Google has released EmbeddingGemma 2, an open, lightweight multimodal embedding model announced on the company's developer blog. The model is designed to convert text and other inputs into embeddings for search and retrieval tasks while remaining small enough to run on modest hardware. Developers are discussing the release, with attention on its open availability and what a small multimodal embedding model means for building search, RAG, and classification applications without heavy compute.