Google DeepMind releases EmbeddingGemma 2 multimodal embedding model

Google DeepMind released EmbeddingGemma 2, a 740-million-parameter open model under Apache 2.0 that natively maps text, code, images, video, and audio into a unified embedding space. It is modular (down to 270M for text/code), optimized for on-device use with quantization, and features an 8K context window. Benchmarks show gains especially on code retrieval compared to the prior version.