DeepSeek V4 Flash Weekly Calls Hit 7.22T Tokens, Top 5 Models Are Chinese

DeepSeek V4 Flash leads global AI usage with 7.22T tokens weekly. Chinese models hold 4 of top 5 spots, surpassing US volume for 14 straight weeks.

Woofun AI data shows that DeepSeek V4 Flash achieved the highest global AI model usage between July 27 and August 2, processing 7.22 trillion tokens. On August 1 alone, the model handled 8 trillion tokens via OpenCode. Chinese models occupied four of the top five positions: Xiaomi’s MiMo-V2.5 (5.1 trillion), Tencent’s Hunyuan Hy3 (5.01 trillion), and another DeepSeek V4 Flash variant (3.45 trillion) ranked second through fourth. OpenAI’s GPT-5.6 Luna placed fifth with 2.99 trillion tokens, reflecting a 738% month-on-month increase.

Total weekly call volume for Chinese large models reached 28.13 trillion tokens, exceeding US counterparts for the 14th consecutive week. Kimi K3 dropped to 12th place, falling out of the top ten. Free models are gaining market share, with some seeing over 50% monthly growth. Domestic providers are prioritizing inference performance and service stability over price competition, while OpenAI and Anthropic continue rapid iteration to maintain user retention.