Google DeepMind releases EmbeddingGemma 2, an open multimodal embedding model

Google DeepMind has released EmbeddingGemma 2, a 740-million-parameter model designed to create embeddings across multiple data types. The model can work with text, code, images, video, and audio, mapping them into a shared embedding space for comparison and retrieval tasks.

The model is built for on-device inference, making it practical for applications with latency or privacy constraints. Google released it under an Apache 2.0 license, making it freely available for commercial and research use.