SignalTech/AI
EmbeddingGemma 2: an open, lightweight multimodal embedding model
New open lightweight multimodal embedding model signals continued AI model release activity.
Context
What happened: Google launched EmbeddingGemma 2, an open multimodal embedding model built on the Gemma 4 architecture and released under an Apache 2.0 license. It unifies code, images, video, and audio in a shared embedding space.
Key numbers: 740 million parameters. 20 million downloads of EmbeddingGemma 1.
Why it matters: The model targets on-device inference, letting developers build cross-modal search and retrieval that works offline. Because it shares Gemma 4's text tokenizer and audio encoder, both models can run in a unified pipeline with a lower combined memory footprint.
What changed
Google launched EmbeddingGemma 2, an open multimodal embedding model built on the Gemma 4 architecture and released under an Apache 2.0 license. It unifies code, images, video, and audio in a shared embedding space.
Why it matters
The model targets on-device inference, letting developers build cross-modal search and retrieval that works offline. Because it shares Gemma 4's text tokenizer and audio encoder, both models can run in a unified pipeline with a lower combined memory footprint.