Meituan has officially unveiled and open-sourced its maiden large-scale model, LongCat-Flash-Chat. This groundbreaking model employs an innovative hybrid expert architecture, boasting a formidable 560 billion parameters, enabling efficient AI via dynamic computation. LongCat-Flash-Chat has demonstrated exceptional performance across multiple benchmark tests, particularly in agent tasks, and boasts ultra-fast inference speeds, rendering it ideal for sophisticated applications. This open-source endeavor has ignited vigorous discussions within both domestic and international technology communities.