NextFin News -- Xiaomi has announced the launch of its large model, Xiaomi MiMo-V2-Pro, which is  designed for high-intensity agent working scenarios in the real world. It boasts over 1 trillion total parameters (42 billion active parameters) and employs an innovative hybrid attention architecture, supporting an ultra-long context length of 1 million tokens. The MiMo-V2-Pro model is now officially available with API services, supporting a context length of 1 million tokens and offering tiered pricing based on usage. For contexts within 256,000 tokens: input is priced at $1 per million tokens and output at $3 per million tokens; for contexts within 1 million tokens: input is priced at $2 per million tokens and output at $6 per million tokens. MiMo-V2-Pro has been launched simultaneously on multiple platforms.