Samsung Gauss2 Generative artificial intelligence model

Main article: Generative artificial intelligence

2024: Samsung Gauss2 Announcement

Samsung Electronics at the end of November 2024 introduced the Samsung Gauss2, the second generation of its own AI model. The solution differs, according to the developers, in optimized performance and various applications.

Samsung Electronics is engaged in software development, including AI and data analytics. Samsung Gauss2, introduced by three different models, already boosts our internal performance, and we plan to integrate it into other products as well. narrated by Paul Kyungwoon Chun, President, CTO of Device eXperience (DX) and Head of Samsung Research

𝓲

Samsung Electronics

Samsung Gauss2 is an optimized generative AI model introduced by the company in 2023. It is characterized by optimized performance and efficiency while simultaneously processing various types of data, since it is a multimodal model that combines language, code and images. It is available in three different models designed for different purposes: Compact, Balanced and Supreme.

The Compact model is a compact model specifically designed to work even in limited computing environments. It provides optimized performance when working with a device, making the most of its computing resources. The Balanced model aims to achieve a balance between performance, speed and efficiency, ensuring stable performance for various tasks. Finally, Supreme aims to achieve optimal performance through the use of Mixture of Experts technology in the Balanced model, which reduces computational costs during training and inference, while maintaining the optimal level of performance and efficiency.

Samsung Gauss2 supports from 9 to 14 languages, as well as various programming languages, depending on the model. Samsung developed and implemented its own Large Language Model (LLM) processing methods and built its own tokenizer to provide efficiency for supported languages. The Balanced and Supreme models have equal to or superior performance in key parameters such as generating responses in various languages, including English and Korean, and coding capabilities compared to leading open source generative AI models. In addition, the data processing speed per hour is 1.5-3 times faster. This allows you to generate AI responses faster, reduce user wait times and optimize task processing efficiency.

The presence of an internal AI generative model makes it easier to customize the system to achieve optimal performance for specific purposes. Samsung Gauss is already being used to solve various problems by Samsung employees who use its flexible development capabilities.

With Samsung Gauss, software developers are assisted by their own code.i coding assistant. For November 2024, it has been updated to Samsung Gauss2 and is used by units of the Device eXperience (DX) division and foreign research institutes.

Since launching in December 2023, the monthly use of code.i has quadrupled and is now used by about 60% of all software developers in the Device eXperience (DX) division. In addition, Samsung Gauss Portal is an artificial intelligence service based on Samsung Gauss that helps employees of the Device eXperience (DX) division solve various office tasks, such as document synthesis, translation and writing of emails. In April 2024, this service was expanded to foreign branches. In addition, starting in August, Samsung is using Samsung Gauss for call centers employees to automatically classify and summarize customer calls.