UI-TARS, developed by ByteDance's research team, is a next-generation native GUI proxy model aimed at seamless interaction with graphical user interfaces leveraging human-like perception, reasoning, and action capabilities. This model integrates all key components such as perception, reasoning, localization, and memory, enabling end-to-end task automation without predefined workflows or manual rules. Its main advantages include powerful multi-modal interaction capabilities, high-precision visual perception and semantic understanding, and excellent performance across various complex task scenarios. This model is particularly suitable for automation of GUI interactions, such as in automated testing and smart office environments, significantly improving work efficiency.