Xiaomi MiMo-V2.5 large model appeared on CCTV and ranks first in the world in terms of calls

Kuai Technology reported on August 2 that according to CCTV News, the global multi-model aggregation platform OpenRouter recently announced the latest weekly AI large model call volume list. The top five models were all developed by Chinese companies. Among them, Xiaomi MiMo-V2.5 topped the list with a call volume of 10.5 trillion Tokens in a single week, a month-on-month increase of 12%. It is understood that Xiaomi MiMo-V2.5 officially launched public beta on April 23 this year, and completed the full series of open source by the end of April. As the volume of model calls continues to grow, technicians from Xiaomi Group's AI laboratory are conducting a new round of system debugging in response to the rapidly rising traffic. Xin Yanshu, product manager of Xiaomi's MiMo large model team, said that MiMo-V2.5's single-week call volume has increased from 1.5 trillion Tokens to 10.5 trillion Tokens, an increase of more than 6 times in two months, and ranked first in the OpenRouter weekly list and monthly list. In addition, in June this year, Xiaomi and TileRT launched MiMo-V2.5-Pro-UltraSpeed, achieving a breakthrough in model inference speed. According to reports, this is the world's first trillion-parameter model whose inference speed exceeds 1,000 tokens/s on a general-purpose GPU, setting a new record for the world's fastest inference speed for flagship models.