End-side multimodal modeling

June 27 - Technology media outlet NeoWin published a blog post today (June 27) reporting that following a preview release at the I/O Developer Conference 2025, Google has officially launched the Gemma 3n end-side multimodal model, which supports running locally on cell phones, tablets and laptops and handling multiple data types, including audio, text, images and video. Compared to the preview release in May, the latest full release of Gemma 3n further improves performance by supporting native operation on hardware with 2GB of RAM, with a focus on improving...

May 21 - Today at the I/O 2025 developer conference, Google unveiled the Gemma 3n end-side multimodal AI model, which can run natively on phones, tablets and laptops and process multiple data types, including audio, text, images and video. Gemma 3n runs on devices with less than 2GB of RAM and utilizes the same architecture as the Gemini Nano, but with better performance, revealed Gus Martins, Gemma product manager. He emphasized: "The Gemma 3n is a great way to set...