On the 21st, at the ‘Samsung Developer Conference Korea 2024’, Lee Joo-hyung, Vice President of Samsung Research at Samsung Electronics, introduced ‘Samsung Gauss2’. (Samsung Electronics)

<이미지를 클릭하시면 크게 보실 수 있습니다>

Samsung Electronics has introduced ‘Samsung Gauss2,’ the successor to its generative artificial intelligence (AI) model launched in 2023. This new model processes various data types - text, images, audio, and video - simultaneously, emulating human sensory integration.

The company showcased Samsung Gauss2 during the online ‘Samsung Developer Conference Korea 2024’ on Thursday. This annual event fosters communication and collaboration with software developers.

Samsung Gauss2 is a multimodal model integrating language, code, and image processing. Multimodal AI considers multiple data types together, learning and processing their interrelationships, and the model is available in three versions: Compact, Balanced, and Supreme.

The Compact version is designed for efficient operation in limited computing environments, optimizing performance for on-device applications. The Balanced version offers a cloud-based model that balances performance and speed with stability and efficiency. The Supreme version targets maximum performance for high-demand applications.

Depending on the version, Samsung Gauss2 supports 9 to 14 languages and various programming languages. Samsung has developed and applied its own large language model (LLM) training stabilization techniques. According to the company, the Balanced and Supreme models perform on par with or surpass existing generative AI models in key metrics, including language support and coding capabilities.