From the source
Google DeepMind released EmbeddingGemma 2, a 740-million-parameter multimodal embedding model that maps text, images, audio and video into a single embedding space for on-device use.
- The model, built on the Gemma 4 architecture and released under an Apache 2.0 license, achieves leading scores among sub-1B multimodal embedders on benchmarks including MTEB Code and MAEB.
- It requires as little as 191MB of active RAM for text-only weights on a Google Pixel 11 Pro, and supports an 8K token context window that can process up to 5.5 minutes of audio or 29 images.
- Developers can download the weights from
Hugging Face and Kaggle, and deploy via Google AI Edge MediaPipe, LiteRT, or tools such as transformers.js and llama.cpp.




