On March 19th, Xiaomi made an official announcement regarding the release of its advanced large model, the Xiaomi MiMo-V2-Pro. This model has been meticulously crafted to excel in demanding Agent work environments. It boasts an impressive total of over 1 trillion parameters (with 42 billion actively engaged parameters), incorporates an innovative hybrid attention mechanism, and is capable of handling an ultra-long context of up to 1 million tokens. At present, MiMo-V2-Pro has initiated its API services, accommodating a context length of 1 million tokens. It employs a tiered pricing strategy: for contexts not exceeding 256,000 tokens, the input fee is set at $1 per million tokens, while the output fee stands at $3; for contexts within the 1 million token range, the input fee is $2 per million tokens, and the output fee is $6. This cutting-edge model has been concurrently launched across multiple platforms.
