Samsung’s latest genAI model, Gauss 2, focuses on internal productivity and future integration into consumer devices.
Samsung’s Gauss 2, an upgrade from the original Gauss, offers three versions: compact, balanced, and supreme. Each caters to different needs, from on-device efficiency to high-performance applications. Following the success of its original Gauss model, Samsung has unveiled Gauss 2 at its developer conference. This new model is not yet integrated into consumer products like the Galaxy S24 Ultra. Instead, its initial focus is on internal applications designed to streamline various tasks and assist employees. Samsung aims to leverage Gauss 2 to enhance user experience through AI and data analytics.
Gauss 2 offers greater efficiency and power compared to its predecessor. The compact model serves on-device applications with limited computational resources. The balanced model optimizes performance and efficiency, while the supreme model boasts high performance and reduced computational costs during both training and execution. This flexibility allows Samsung to approach various use cases effectively.
As a multimodal model, Gauss 2 supports 9 to 14 languages, including English and Korean. Samsung’s custom tokenizer and stabilization techniques improve efficiency when training large language models. This capability positions Gauss 2 to cater to diverse tasks, such as document summarization, translation, and email composition.
Internally, Gauss 2 has dramatically boosted productivity. Usage of code.i, an internal tool, has quadrupled since the model’s launch, with 60% of developers in the Device eXperience Division actively using it. Although there are speculations about integrating Gauss 2 into Samsung’s future Android 15-based One UI 7, the company has yet to confirm these plans. Overall, Gauss 2 exemplifies Samsung’s commitment to harnessing generative AI for improved operational efficiency.