Skip to content
by InfinyAI ES

IA en un minutoNewsModels

ModelsUpdated:

Google launches EmbeddingGemma 2, an open model that links text, images, audio and video on the device itself

IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds

EmbeddingGemma 2 is Google's new open model for search that no longer stops at words: it places text, images, video, audio and code in a single vector space. With 740 million parameters, it builds on Gemma 4, and its Apache 2.0 license lets companies use it in commercial products. It can be downloaded from Hugging Face and Kaggle, with Model Garden to follow.

Google is taking its model for searching information on the device itself beyond text.

It has launched EmbeddingGemma 2, an open embedding model (it turns content into vectors so it can be searched and connected) that brings text, code, images, audio and video together in a shared space. It has 740 million parameters, is built on the Gemma 4 architecture (Google's generative model) and is released under the Apache 2.0 license, which allows commercial use.

According to Google, it can help find a video clip from a voice memo or search through hours of audio recordings with a text query. The new model is already available on Hugging Face and Kaggle (model platforms), and it is coming soon to Model Garden, Google's model catalog for businesses.

Google launched the first EmbeddingGemma last year, for text only, and it has passed 20 million downloads. The new one is modular: text-only work needs as little as 270 million parameters, and the vision (170 million) and audio (300 million) encoders are optional. Quantized, on a Google Pixel 11 Pro it needs about 191 MB of RAM for text and about 567 MB for the full multimodal model.

Its context window is 8,000 tokens, four times that of the first model, which lets it process up to 5.5 minutes of audio, 29 images or 58 video frames at once. On code, its MTEB Code score rises from 68.76 to 78.68.

Why it matters · analysis and opinion

The first EmbeddingGemma only understood text, which left out much of what a company keeps: meeting recordings, product photos, training videos. The new version brings all of that into the same search and, most strikingly, inside the phone or computer itself, without sending files to any server. The first to notice will be those building search apps and assistants that work offline, and companies that cannot upload their material to the cloud for privacy reasons. What still has no date is its arrival in Google's catalog for businesses.

Official source: Google · Written with the help of AI: how we make the news

AI in your inbox, every day or every Friday

The stories that matter, each one in a minute. With the source for every one.

Choose one or both:

Sign up and you are in: the daily arrives every night and the weekly on Friday mornings. You can unsubscribe from any email.