Google DeepMind DeepMind
code
Build with Gemini
spark
Chat with Gemini

EmbeddingGemma

A best-in-class text embedding model optimized for on-device use cases.

Explore EmbeddingGemma

keyboard_arrow_right
Read docs
keyboard_arrow_right
View tech report

EmbeddingGemma generates high-quality embeddings with reduced resource consumption, enabling on-device Retrieval Augmented Generation (RAG) pipelines, semantic search, and generative AI applications that can run on everyday devices.

Capabilities

speed

Engineered for efficiency

A 308M parameter model that can run on less than 200MB of RAM with quantization.

translate

Strong multilingual performance

Trained on over 100 languages, providing best-in-class text understanding for its size.

animation

Fast and flexible

Leverages Matryoshka Representation Learning (MRL) for customizable embedding dimensions.

Performance

EmbeddingGemma is the highest ranking open multilingual text embedding model under 500M parameters on the Massive Text Embedding Benchmark (MTEB).

Scatter plot titled 'MTEB (Multilingual, v2), Score by model size' comparing embedding models. The 'EmbeddingGemma' model is highlighted with a blue dot, showing a mean task score of approximately 61 at a model size of roughly 300M, outperforming similarly sized models like 'gte-multilingual-base' Scatter plot titled 'MTEB (Multilingual, v2), Score by model size' comparing embedding models. The 'EmbeddingGemma' model is highlighted with a blue dot, showing a mean task score of approximately 61 at a model size of roughly 300M, outperforming similarly sized models like 'gte-multilingual-base'

Download EmbeddingGemma

Get started with EmbeddingGemma

Try the model by generating embeddings in an interactive notebook.

keyboard_arrow_right
Run in Colab