We launched EmbeddingGemma final yr to offer a light-weight choice for high-quality textual content embeddings, to assist your apps manage, search, and join info instantly on client {hardware}. The developer group’s response blew previous our expectations. With greater than 20 million downloads, builders have used it to energy smarter on-device search instruments and privacy-first retrieval augmented era (RAG) pipelines.
Right this moment, we’re launching EmbeddingGemma 2, increasing past textual content to unify code, photos, video, and audio in a shared embedding house. Constructed on the Gemma 4 structure and launched underneath a commercially permissive Apache 2.0 license, EmbeddingGemma 2 has 740 million parameters, making it optimum for on-device inference. It could assist discover a particular video clip from a voice memo, or search via hours of audio recordings primarily based on a textual content question, all processed by a single, natively multimodal mannequin.
