A trillion-parameter scale MOE-32K long text model. Achieves absolute leading levels across various benchmarks, capable of handling complex instructions and reasoning, with advanced mathematical abilities, supporting function calls, and optimized for applications in multilingual translation, finance, law, and healthcare.