Google Releases EmbeddingGemma 2, 740M-Parameter Multimodal Model for On-Device Semantic Search

Google DeepMind has released EmbeddingGemma 2, a 740-million-parameter embedding model that converts text, images, video, and audio into a unified vector space for local semantic search and retrieval-augmented generation (RAG) on edge devices. Released October 6, 2026, under an Apache 2.0 license, the model represents a significant expansion of Google's open EmbeddingGemma line, which has accumulated over 20 million downloads since its September 2025 launch.