[China's Large AI Models Lead Weekly Usage for Five Consecutive Weeks: Xiaomi MiMo-V2.5 Surges Over 999% Week-on-Week, StepStar Step 3.5 Flash Drops Off the List] Jin10 News, June 1 — According to the latest data estimates from OpenRouter, global AI large model usage reached 31.8 trillion tokens last week (May 25 to 31), a 10% increase from the previous week, marking six consecutive weeks of growth. The demand for large model usage continues to expand. Among the AI large models on the leaderboard, China's AI large models recorded a weekly usage of 11.13 trillion tokens, a 20.68% week-on-week increase, achieving growth for two consecutive weeks. During the same period, U.S. AI large models recorded a weekly usage of 4.24 trillion tokens, a 14% week-on-week decline. China's large model usage has surpassed the U.S. for five consecutive weeks, firmly holding the top global position. Chinese models are increasingly integrating into the global developer usage ecosystem. Among the top three global models by usage, the top two are Chinese AI large models: DeepSeek-V4-Flash and Tencent Hy3 preview. DeepSeek-V4-Flash ranked first for the second consecutive week, with a weekly usage of 3.11 trillion tokens, a 9% week-on-week decline. Hy3 preview ranked second for the second consecutive week, with a weekly usage of 3.03 trillion tokens, a 1% week-on-week decline.