B · Normal
[Citi: AI inference efficiency improves and cutting-edge proprietary model task costs drop significantly] On September 28, Citi released a research report stating that the latest cutting-edge models have improved in performance and cost per task, highlighting the improvement in model efficiency and inference economic benefits, which may help strengthen the longer-term economic advantages of higher-efficiency models over lower-efficiency open source weight alternatives. The bank said that the cost per task of the open source model fell 35% week-on-week to US$0.8, and the discount relative to the closed model expanded from 40% to 60%; the intelligence lead of the proprietary model increased from 9 points to 12 points.
Comments