According to the latest data from OpenRouter as of this week, DeepSeek-V4-Flash has officially topped OpenRouter's global ranking of AI model usage, leaving all other competitors far behind with a weekly invocation volume of 3.43 trillion tokens.
From May 18 to May 24, the total global invocation volume of AI models reached 28.9 trillion tokens, representing a 7.4% increase from the previous week and marking the fifth consecutive week of growth.
The surge in AI agent applications is one of the most direct drivers behind this increase. Autonomous task execution, multi-step reasoning, and external tool invocation consume hundreds of billions of tokens. At the same time, the large-scale implementation of use cases such as code generation, long-document processing, and enterprise knowledge base retrieval continues to drive up the invocation volume of foundational models.
Among the AI models ranked last week, Chinese models recorded a weekly invocation volume of 9.223 trillion tokens, a 19.89% increase month-over-month. Over the same period, US models saw an invocation volume of 4.93 trillion tokens, a 16.27% increase. Chinese large models have now surpassed their US counterparts in weekly invocation volume for four consecutive weeks, firmly holding the top position globally.
Related News:
Frontier | CEO says AI will replace 'Measurers' as Cloudflare lays off 20%
Frontier | Meta employees rush for AirPods before 8,000 job cuts hit
Comment