DeepSeek V4 Flash ranked first in global AI large model usage for the week of July 27 to August 2, according to data from multi-model aggregation platform OpenRouter. The model recorded 7.22 trillion tokens in call volume.
OpenCode data shows the model handled 8 trillion tokens on August 1 alone. Chinese models occupied four of the top five spots: Xiaomi MiMo-V2.5 (5.1 trillion), Tencent Hunyuan Hy3 (5.01 trillion), and another DeepSeek V4 Flash (3.45 trillion) placed second through fourth, followed by OpenAI GPT-5.6 Luna at fifth with 2.99 trillion tokens, up 738% week over week.
China's total large model call volume for the week reached 28.13 trillion tokens, marking the 14th consecutive week above the U.S. Kimi K3, which had drawn attention earlier, dropped out of the top ten to 12th place.
Free models are capturing a growing share of traffic, with some models growing over 50% week over week. Analysts say domestic models have shifted from low-price competition toward reasoning performance and stable service capabilities. With OpenAI and Anthropic continuing to iterate, sector momentum is rotating faster, and product iteration speed will determine user retention.