Understand this story
On November 17, 2026, Google released EmbeddingGemma 2, expanding the open-source embedding model that previously supported only text into a natively multimodal one that can map text, images, audio, video, and any combination thereof into the same embedding space, while also compressing the model backbone to be smaller. Google simultaneously released a technical report disclosing the training strategy, evaluation methods, and experimental results, and stated that its approach is based on technical innovations from Gemma 4 and Gemini Embedding 2. There are currently no subsequent developments.
AI-generated from public reports
Latest development · Report
Google Releases EmbeddingGemma 2 Technical Report, Extending Embedding Model to MultimodalDevelopments
1 confirmed developments
- Source:Google Blog
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 17
Google BlogFirst-partySelected
Google Releases EmbeddingGemma 2 Technical Report, Extending Embedding Model to Multimodal
Google has released EmbeddingGemma 2, extending its open-source embedding model that previously supported only text into a natively multimodal one that can map text, images, audio, video, and any combination thereof into the same embedding space, while also compressing the model backbone to be smaller. Google simultaneously released a technical report disclosing the training strategy, evaluation methods, and experimental results, and stated that its approach builds on technical innovations from Gemma 4 and Gemini Embedding 2.
AI Trends