Google EmbeddingGemma 2

Google has released EmbeddingGemma 2, a lightweight multimodal embedding model that can process text, code, images, video, and audio into a unified embedding space with a 740M parameter size and 8K context window. The model is optimized for on-device use and offers flexible output dimensions through Matryoshka Representation Learning, available under an Apache 2.0 license.