On June 9, the GN06 team under Meituan officially launched the AI-native browser Tabbit 1.0. As the core entry point of the Tabbit browser, this product marks the transition of this intelligent assistant deeply integrated with large models from the internal testing phase to large-scale commercial application. Since its test launch in March of this year, the task completion rate of Tabbit Agent has seen a significant increase, rising from an initial 53.1% to 90.8% at present, achieving industry-leading levels in interaction accuracy and execution stability.
In terms of technical core, Tabbit 1.0 is based on the self-developed LongCat large model, and integrates more than ten mainstream models such as DeepSeek, Kimi, and Tongyi Qianwen, building a powerful "model family" ecosystem. The product is capable of accurately executing complex UI operations such as clicks, scrolling, and cross-website collaboration, compressing the average response time for basic tasks to within 10 seconds. This update not only enhances the view-sharing function of the AI conversation in the sidebar but also achieves full-platform coverage on Windows, macOS, and mobile devices (iOS/Android), ensuring a consistent smart office experience across different devices.
Liu Jiong, head of the GN06 team, stated that the mission of Tabbit is to lower the barrier to using AI technology. Through the "Magic Tips Square" and creator support program, the browser has evolved from a simple information carrier into an efficient AI work entrance. Industry observers believe that the launch of Tabbit 1.0 represents a key turning point in the transformation of AI Agents from "dialogue-based consultation" to "action-based delivery." With the significant improvement in agent execution efficiency, the browser is becoming the core foundation of personal digital productivity and is expected to completely reshape the future human-computer interaction paradigm and office workflow logic.
