Samsung Gauss2 multimodal AI model with support for 14 languages ​​unveiled at SDC24

Samsung on Thursday unveiled the second generation of its Gauss artificial intelligence (AI) model. The new multimodal AI model called Gauss2 is said to facilitate advanced use cases for improved performance, efficiency, and application integration. The new large language model (LLM) was introduced during the keynote address at the Samsung Developer Conference Korea 2024 (SDC24 Korea) which was held online. The tech giant highlighted that it is using the capabilities of AI models in software development and research purposes.

Samsung Gauss2 AI model unveiled

In a newsroom post, the South Korean tech giant detailed the second generation of its Gauss AI model, which was launched last year. The new version of the foundation model now comes with several upgrades. The company claimed that multimodal capability, which involves handling data sets across different modalities, has been improved.

Additionally, AI models now support 9 to 14 languages ​​as well as different programming languages. This is said to provide better performance across language, code and images.

Samsung Gauss2 is available in three different models based on their parameter sizes – Compact, Balanced and Supreme. Compact is a small-sized model designed for efficiency and working in limited compute environments. The Balanced Model is optimized for both performance and efficiency, and the Supreme Model can handle high-level computing tasks by leveraging Mixture of Experts (MOE) technology.

Samsung claims that the Balanced and Supreme models can outperform “leading open-source generative AI models” in English and Korean language-based tasks, as well as coding-related tasks. The company also claims that the models provide lower latency times, faster processing speeds, and better task management compared to open-source models.

Samsung Gauss2 is currently being used in the company’s Device Experience (DX) division and overseas research institutes. The most popular use case of AI models is an internal coding assistant called Code.ai that helps with software development. The tech giant claimed that 60 percent of its DX division uses this tool. This technology is also being used by company call center staff to classify and summarize customer calls.

While Gauss2 is currently being used internally, the company plans to ship it with its own products as well. Samsung also believes that the AI ​​model can enhance the personalization capabilities of its existing AI features.