China's Large Model Tops Global Weekly Queries, MiniMax Leads the Pack

In the race for AI computing power, China's large language models are achieving a "surpassing turn" at a remarkable pace. According to the latest monitoring data from OpenRouter , from March 2 to March 8, the weekly call volume for China's large models reached 4.19 trillion Tokens, a surge of 34.9% compared to the previous week.
This is more than a simple numerical breakthrough; it marks the second consecutive week that China's large models have surpassed their U.S. counterparts in call volume (U.S. models handled 3.63 trillion Tokens during the same period, an 8.5% decrease). This shift highlights the unprecedented dynamism within China's AI application ecosystem.
Three Chinese Firms in Global Top Five, MiniMax Holds the Lead
In the global rankings for large model call volume, Chinese products now occupy three of the top five positions:
MiniMax M2.5: Maintains its global number one spot with a weekly call volume of 1.87 trillion Tokens, a 15% increase. Its high cost-effectiveness and reliable output have made MiniMax a top choice for developers.
DeepSeek V3.2: Ranks third globally with 0.83 trillion Tokens in weekly calls, showcasing its deep expertise in managing complex tasks.
Stepwise Star Step3.5 Flash: The most notable newcomer this week, with a call volume of 0.75 trillion Tokens, surged by 69% to break into the global top five.
It is worth noting that Kimi K2.5 , previously a strong performer, dropped to sixth place globally this week, though its call volume still saw a steady 8% growth.
A Dual Victory in Compute and Application
The surge in China's large model usage is no coincidence. On one hand, providers like MiniMax and DeepSeek have continuously optimized their technology, significantly lowering the cost per Token and making their models more accessible. On the other hand, China's vast array of end-user scenarios and industry-specific demands provide a fertile testing ground for practical AI applications.
As the global AI competition enters its next phase, the focus is shifting from sheer parameter scale to deeper integration into real-world workflows. China's models consistently outperforming U.S. counterparts may signal a significant shift in the global AI landscape's center of gravity. In the tension between "excess computing power" and "application hunger," China's AI sector is rapidly forging its own path forward.
Related article
Anthropic Expands Claude AI Coding Tools to Japan in Push for Overseas Growth
Anthropic, a prominent U.S. artificial intelligence firm, is intensifying its global outreach. On Wednesday, the company hosted a major developer gathering, "Code with Claude," in Tokyo, drawing close to 500 software engineers. This initiative seeks
India’s Software Giant Cuts Hiring, Promises No Layoffs as AI Agents Scale Up
As artificial intelligence reshapes traditional labor-intensive business models, Tata Consultancy Services (TCS), a premier Indian software outsourcing firm, has unveiled its strategic response. During Tuesday’s annual general meeting, the chairman o
China Locks AI Models During Gaokam to Block Instant Homework Help
With the 2026 Gaokao fast approaching, rumors regarding the suspension of AI tools during the exam period have ignited intense online debate. In response to public concern, major AI platforms and educational apps have clarified their stance: rather t
Related Special Topic Recommendations
Comments (0)
0/500

In the race for AI computing power, China's large language models are achieving a "surpassing turn" at a remarkable pace. According to the latest monitoring data from
This is more than a simple numerical breakthrough; it marks the second consecutive week that China's large models have surpassed their U.S. counterparts in call volume (U.S. models handled 3.63 trillion Tokens during the same period, an 8.5% decrease). This shift highlights the unprecedented dynamism within China's AI application ecosystem.
Three Chinese Firms in Global Top Five, MiniMax Holds the Lead
In the global rankings for large model call volume, Chinese products now occupy three of the top five positions:
MiniMax M2.5: Maintains its global number one spot with a weekly call volume of 1.87 trillion Tokens, a 15% increase. Its high cost-effectiveness and reliable output have made
DeepSeek V3.2: Ranks third globally with 0.83 trillion Tokens in weekly calls, showcasing its deep expertise in managing complex tasks.
Stepwise Star Step3.5 Flash: The most notable newcomer this week, with a call volume of 0.75 trillion Tokens, surged by 69% to break into the global top five.
It is worth noting that
A Dual Victory in Compute and Application
The surge in China's large model usage is no coincidence. On one hand, providers like
As the global AI competition enters its next phase, the focus is shifting from sheer parameter scale to deeper integration into real-world workflows. China's models consistently outperforming U.S. counterparts may signal a significant shift in the global AI landscape's center of gravity. In the tension between "excess computing power" and "application hunger," China's AI sector is rapidly forging its own path forward.
Anthropic Expands Claude AI Coding Tools to Japan in Push for Overseas Growth
Anthropic, a prominent U.S. artificial intelligence firm, is intensifying its global outreach. On Wednesday, the company hosted a major developer gathering, "Code with Claude," in Tokyo, drawing close to 500 software engineers. This initiative seeks
India’s Software Giant Cuts Hiring, Promises No Layoffs as AI Agents Scale Up
As artificial intelligence reshapes traditional labor-intensive business models, Tata Consultancy Services (TCS), a premier Indian software outsourcing firm, has unveiled its strategic response. During Tuesday’s annual general meeting, the chairman o
China Locks AI Models During Gaokam to Block Instant Homework Help
With the 2026 Gaokao fast approaching, rumors regarding the suspension of AI tools during the exam period have ignited intense online debate. In response to public concern, major AI platforms and educational apps have clarified their stance: rather t





Home






