Google has released EmbeddingGemma 2, an open-source, lightweight, and multimodal embedding model. This new model expands beyond text to handle images, video, and audio within a shared embedding space, built on the Gemma 4 architecture. It offers significant improvements in code and multimodal performance while maintaining efficiency for on-device applications, utilizing Matryoshka Representation Learning for reduced storage needs and an 8K token context window. AI
IMPACT Enables more powerful and private on-device AI applications by unifying multimodal data processing.
RANK_REASON New model release from a major AI lab (Google) with significant multimodal capabilities and on-device optimization. [lever_c_demoted from frontier_release: ic=2 ai=1.0]
- Apache Software License 2.0
- EmbeddingGemma
- Gemini Embedding models
- Gemma 4
- Google Pixel 11 Pro
- GPT-4
- Matryoshka Representation Learning
- MTEB
- OpenAI
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
