Google has launched EmbeddingGemma 2, a new open AI model designed to understand different types of information together. Google CEO Sundar Pichai announced the model, calling it the company first open, natively multimodal embedding model. The new model can work with text, code, images, video and audio. Unlike systems that need separate models for different types of data, EmbeddingGemma 2 can bring these inputs into one shared space. This can make search and information retrieval faster and simpler.
Google says the model has 740 million parameters and is built on the Gemma 4 architecture. It is designed to run directly on devices such as phones and laptops, which means some AI tasks can work without sending data to the cloud. This could also improve privacy and reduce delays.
For example, users could search for a particular video using a voice recording or find images using a text description. The model can also help developers build local search systems and AI applications that work with different kinds of media. Google has released EmbeddingGemma 2 under the Apache 2.0 licence, allowing developers to use and build applications with the model. The model weights are also available through platforms including Hugging Face and Kaggle.
