Google unveils updated Gemini Nano AI model for upcoming Galaxy series

Samsung Electronics is in talks with Google to install its latest AI model on next year's Galaxy S25.

Alphabet CEO Sundar Pichai speaks at a Google I/O event in Mountain View, California on Tuesday. [YONHAP]

Google unveiled an updated Gemini Nano for on-device AI that is expected to be installed on Samsung Electronics' upcoming Galaxy S25 smartphones.

The Big Tech company held a developer conference, Google I/O 2024, in California on Tuesday where it laid out key updates and a road map for AI to enter what it calls "the Gemini era."

The updated Gemini Nano, the smallest version of Google's large language model, comes topped with multimodality, able to process not only text but also video, pictures and sound data.

The model will be deployed in Google's Pixel phones in the latter half of the year, according to the tech company.

Samsung Electronics is also in talks to install the latest version of the AI model on its Galaxy S25 smartphones, which are slated for release early next year.

"Talks on how to realize the second version of Gemini Nano, which will have multimodal features for on-device AI, have commenced," said Jo Chul-min, vice president of Samsung Electronics' software development division, at a conference in April.

In order to optimize the model on Samsung's Exynos 2500 processor, which is expected to power the Galaxy S25, the company is planning to expand the related research team by two to threefold, according to Jo.

Speculation was also reignited when Samsung Electronics President Roh Tae-moon, in charge of the mobile business, met with Rick Osterloh, the vice president of Google responsible for platform and device operation, later that month.

"Exciting things are coming up for the future of AI-powered Android and Galaxy experiences," a tweet from Samsung read after the meeting.

The partnership between the two companies has become tight recently.

The Korean IT giant's first AI smartphone, the Galaxy S24, which launched early this year, has partially deployed the first version of the Gemini Nano along with its own Gauss model to run its on-device AI features. It also introduced Google's flashy Circle-to-Search feature for the first time.

At its developer conference, Google also introduced Project Atra, an AI agent that takes on more natural conversation with the user and its own tensor processing unit, Trillium, which has increased the deployment of high bandwidth memory chips twice over compared to the previous version.

BY JIN EUN-SOO [jin.eunsoo@joongang.co.kr]