Google expands EmbeddingGemma beyond text to images, audio and video

What happened

Google LLC today released EmbeddingGemma 2, an open multimodal (able to work with more than one kind of input, such as text and images) embedding model small enough to run on a smartphone. The release takes the EmbeddingGemma line beyond text, which was all the first version handled when Google introduced it in September 2025. Google is an Internet and software company based in Mountain View, and its products and services include Google Voice.

An app built on […] The post Google expands EmbeddingGemma beyond text to images, audio and video appeared first on SiliconANGLE . UPDATED 18:18 EDT / OCTOBER 06 2026 AI Google expands EmbeddingGemma beyond text to images, audio and video by Duncan Riley Google LLC today released EmbeddingGemma 2, an open multimodal embedding model small enough to run on a smartphone.

The release takes the EmbeddingGemma line beyond text, which was all the first version handled when Google introduced it in September 2025 . Built on the Gemma 4 architecture Google released in April , the new version is more than twice the size of the original at 740 million parameters.

On its own, the 270 million-parameter text core used about 191 megabytes of memory when Google tested a quantized build on a Google Pixel 11 Pro. An app built on the new model could take a voice memo, for example, and find the matching moment in a video without the data ever leaving the phone.

Most of the growth is in the vision and audio encoders, which apps that only work with text can leave off. An app also has to store what the model produces.

Key facts

• Google LLC today — released: EmbeddingGemma 2, an open multimodal embedding model small enough to run on a smartphone
• Built on the Gemma 4 architecture Google — released: in April , the new version is more than twice the size of the original at 740 million parameters