Wang Dong, co-founder of Moore Threads, said: "There is no 'universal chip' in the market, but rather a combination of solutions."

Founder and CEO of GPU manufacturer Moore Threads, Wang Dong, said, "The development of large models is very rapid both domestically and internationally. Currently, top manufacturers complete a front-end basic model iteration every two months on average. In terms of model invocation costs, we have found that Chinese front-end basic models have significant cost advantages compared to equivalent foreign models with the same intelligence level. Chinese models have better cost performance, which precisely indicates that model companies have done a lot of work in improving model efficiency, cost efficiency, and training costs under limited computing power." Wang Dong pointed out that the inference market does not have a "universal chip," but rather a combination of "solutions." "The technical application threshold in the inference market is relatively low, and the scenario is highly fragmented. There is no company that can monopolize all segmented application scenarios. There is no absolutely perfect single hardware. Through flexible software and hardware coordination, each model can find the most suitable hardware combination for it to achieve the best balance between cost and performance. The market will see a large number of ISP companies emerging to provide more cost-effective and flexible customized inference services for MaaS providers or end customers."