EmbeddingGemma 2 brings multimodal embeddings to local devices
Google DeepMind released EmbeddingGemma 2, an open embedding model for text and code, images, video, and audio. Its modular architecture has about 740M parameters and supports a shared 8,192-token input budget.
Access at announcement
Public weights are available on Hugging Face and Kaggle under Apache 2.0; the model card also requires adherence to the Gemma Prohibited Use Policy. Model Garden availability is forthcoming. Media modalities share the same input budget.
Sources for this release3 sources
Related release · EmbeddingGemma 2 brings multimodal embeddings to local devices
Related release · EmbeddingGemma 2 brings multimodal embeddings to local devices
Related release · EmbeddingGemma 2 brings multimodal embeddings to local devices
Collection scope
Release links and snapshots containing the exact model are included here. Family background is counted separately.
Model family catalog1
Current models
Official catalogEmbeddingGemma 2Available8.192K ContextOpen weights
- Model ID
google/embeddinggemma-2- Released
- 2026-10-06
- Availability
- Available
- Input
- Text · Image · Audio · Video
- Output
- Embedding
- Context
- 8,192 tokens
- Parameters
- 0.744371512B
- Deployment / access
- Open weights · Local
- Weights
- Open weights
- License
- Apache 2.0
- Capabilities
- Multilingual
Model card rounds size to 740M; the official HF BF16 checkpoint reports 744,371,512 parameters. Native output is 768 dimensions, truncated to 512/256/128; 128 is mainly for text. All modalities share 8,192 tokens. Local hardware requirements vary; Model Garden is forthcoming. Card also requires the Gemma Prohibited Use Policy.