On August 26th, insider sources from Sugon revealed that the company is set to launch a cutting-edge token acceleration solution, tailored specifically for large-scale model inference. This new offering is designed to tackle the issues of redundant computation and inefficient data scheduling that often plague inference processes. At present, Sugon's ParaStor distributed storage system has already been implemented in China's pioneering, fully domestically-produced AI supercomputing cluster, the Sugon 8000 (Dengfeng), which boasts an impressive 100,000 cards. This deployment provides robust data throughput and flexible scalability, catering to the demands of large-scale computing tasks. As per the latest report from CCID Consulting, Sugon has secured the top spot in four key domains: AI, education, embodied intelligence, and autonomous driving.