Google DeepMind releases EmbeddingGemma 2, an open multimodal embedding model
Google DeepMind has announced the release of EmbeddingGemma 2, an open and lightweight multimodal embedding model. The company states that EmbeddingGemma 2 is designed for on-device semantic search, capable of mapping five different modalities—text, image, audio, video, and 3D data—into a single embedding space. Google claims the model is "best-in-class" and outperforms rival embedding models that are twice its size. This release and its associated performance claims are primarily reported by Google's own blogs and community posts, with other outlets echoing these details.
11 articles from 8 outlets covered this story. Their coverage differs on 1 point. The underlying claim is sourced from a shipped.
What do all outlets agree on?
8 outlets covered “Google DeepMind releases EmbeddingGemma 2, an open multimodal embedding model”. All of them report the following:
- Google DeepMind released EmbeddingGemma 2
- It is an open model
- It is a multimodal embedding model
- It is designed for on-device/edge semantic search
- It supports five modalities (text, image, audio, video, 3D)
- Google claims it is "best-in-class" and outperforms larger rivals
Did outlets disagree about this?
Yes. Coverage of “Google DeepMind releases EmbeddingGemma 2, an open multimodal embedding model” differs on 1 point. Each account below is how a different outlet described the same event:
EmbeddingGemma 2 is built on Gemma 4
Which outlets covered this?
All 11 articles found on this story, grouped by the stance of the piece. Every link goes to the original publisher.