Introducing EmbeddingGemma 2: An open model for natively multimodal embeddings
Google releases EmbeddingGemma 2, a sub-billion open model unifying text, image, video, and audio embeddings on-device.
“a picture of a cat, the word cat, and the sound of a cat are all mapped close to each other in a shared high-dimensional embedding space”
Google launched EmbeddingGemma 2, an open model with up to 740M parameters that maps text, images, video, and audio into a single unified embedding space, setting a new benchmark for sub-billion parameter models. It runs entirely on-device with no API calls required, enabling private multimodal search and RAG pipelines where sensitive data never leaves the hardware. The model supports Matryoshka representation learning for flexible output dimensions and can be fine-tuned for domain-specific use cases like legal, medical, or product catalog retrieval.