Google DeepMind launches EmbeddingGemma 2, a 740M-parameter model to map code, images, video, and audio in a shared embedding space, under an Apache 2.0 license (Google)

Google DeepMind’s EmbeddingGemma 2 brings on-device multimodal embeddings to a smaller 740M-parameter model, enabling organizations to unify text, code, images, video, and audio in a shared representation without sending sensitive data to the cloud. For CIOs, the strategic value is lower latency, better privacy, and reduced inference costs for search, retrieval, personalization, and agentic workflows at the edge, while the Apache 2.0 license lowers adoption friction and expands experimentation. IT teams should view this as a building block for more scalable multimodal applications and an opportunity to standardize embedding infrastructure across products and internal platforms.