China's AI Large Model API Calls Lead Globally for Nine Consecutive Weeks as US Share Plunges from 72% to 33%

Chinese AI large models have maintained the global lead in API call volume for nine consecutive weeks, according to the latest OpenRouter data tracked by National Business Daily. During the week of June 22-28, total global AI model calls reached 46.7 trillion tokens, with Chinese models accounting for 20.39 trillion tokens—an 8.4% increase week-over-week—while US models fell to just 4.25 trillion tokens, a 26.22% decline. The trend reflects a profound shift in the global AI landscape. A report from research firm ExponentialView titled "2026 AI Economy Status" highlights that developers and enterprises are increasingly turning to Chinese models such as DeepSeek, GLM, and MiniMax for cost-effective inference. OpenRouter data reveals that the share of token requests going to US models from Google, OpenAI, and Anthropic has collapsed from 72% a year ago to just 33% in June 2026. Among individual models, DeepSeek-V4-Flash held the top spot for the sixth consecutive week with 4.66 trillion tokens in weekly call volume, followed closely by Xiaomi's MiMo-V2.5 at 4.48 trillion tokens (up 14% week-over-week) and MiniMax M3 at 3.74 trillion tokens. Zhipu AI's GLM-5.2 surged into seventh place with 2.11 trillion tokens, posting a 66% week-over-week gain after its open-source release on June 17. GLM-5.2's rise is particularly noteworthy: it scored 51 points on ArtificialAnalysis's global LLM intelligence index, making it the highest-ranked open-source model. The model excels at coding and long-horizon tasks, capable of autonomously running complete large-scale engineering projects over several hours. Notably, Claude Sonnet 4.6 dropped out of the top rankings for the first time in nearly four months, underscoring the competitive pressure on premium US models. As ExponentialView founder Azeem Azhar observes, "You don't always need a Nobel laureate to extract a number from a receipt and put it into a spreadsheet." The data signals a maturing market where cost efficiency and task-appropriate model selection are increasingly driving deployment decisions.