GoKawiilGoogle releases EmbeddingGemma 2, a multimodal on-device embedding model
Google introduced EmbeddingGemma 2, a 740M-parameter embedding model that maps text, code, images, video, and audio into a single unified embedding space. It supports flexible output dimensions from 768 to 128 via Matryoshka Representation Learning, offers an 8K context window—four times larger than the text-only EmbeddingGemma—and is released under the Apache 2.0 license.
GoKawiil's interpretation of the reporting above, not reported fact.
A permissive license and compact size suggest Google is positioning this model for developers building on-device search, retrieval, and recommendation features across multiple data types without cloud dependency. The multimodal unified embedding space could simplify applications that previously needed separate models for text, image, and audio similarity tasks.
- 740M-parameter multimodal embedding model covering text, code, images, video, and audio
- Supports flexible embedding sizes from 768 to 128 dimensions and an 8K context window
- Released under the Apache 2.0 license, allowing commercial use
Source: twitter.com, 2026-10-06
Published there as: “Google EmbeddingGemma 2”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.