Recently, during the GOSIM2025 conference, Tsinghua University, working in tandem with the China Software Testing Center, unveiled the "2025 Large Model Service Performance Ranking." This ranking is the result of a meticulous evaluation carried out by professional teams, who assessed the service performance of various MaaS (Model as a Service) providers. The evaluation focused on crucial metrics like latency, throughput, and reliability.
Notably, in the throughput test for DeepSeek - R1 - 0528, PPIO emerged as the leader, securing a remarkable score of 45.17 tokens per second. Moreover, when it came to model tests involving DeepSeek V3.1 and Kimi - K2 - Instruct, PPIO's performance in terms of both throughput and latency was strong enough to place it among the top five performers.