EmbeddingGemma is Google's family of small open embedding models, built on Gemma, that turn content into vectors for search and retrieval on your own device.
In plain words
An embedding model turns content into a vector, a list of numbers, that apps use for search, retrieval, classification and clustering. The first EmbeddingGemma, based on Gemma 3, had 308 million parameters and read only text. EmbeddingGemma 2 has 740 million: a 270-million-parameter text model plus optional vision and audio encoders. It puts text, images, video and audio into one shared space and covers more than 100 languages. Vectors can be shortened to save storage, and the weights are open under Apache 2.0.
A quick example
Imagine a notes app on a phone. You type "project timeline" and it finds a photo of a whiteboard and a voice memo about deadlines, because all three were turned into vectors on the phone itself.
Where you'll meet it
- Search and RAG that run on phones and laptops instead of a server.
- Grouping or labelling content by meaning, such as classification and clustering.
Related
Sources
- DocsEmbeddingGemma 2: an open, lightweight multimodal embedding model, Google · read 6 Oct 2026
- DocsEmbeddingGemma, Google AI for Developers · read 6 Oct 2026
- Docsgoogle/embeddinggemma-2, Google (Hugging Face) · read 6 Oct 2026
Checked by WhatIsAI against the sources above · 6 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.