Select Language:
According to recent data from OpenRouter, China’s large-scale language models have experienced significant growth in weekly usage, surpassing American counterparts for the second consecutive week. Between March 2 and March 8, the total number of tokens processed by Chinese large models reached an impressive 41.9 trillion, marking a 34.9% increase compared to the previous week.
Meanwhile, the United States’ large model call volume for the same period was around 36.3 trillion tokens, showing a slight decrease of 8.5% week-over-week. This trend highlights a notable shift in global AI model usage, with China maintaining its lead in recent weeks.
Domestically, Chinese AI products now dominate the top five spots in terms of call volume. Notably, MiniMax’s M2.5 model remains at the top globally, handling a weekly volume of 18.7 trillion tokens—up 15% from the previous week. Following closely is DeepSeek V3.2, which processed approximately 8.3 trillion tokens, reflecting a modest growth of 4%.
Another standout performer was Step3.5 Flash from Jianyue Xingchen, which saw an extraordinary week-over-week increase of 69%, reaching 7.5 trillion tokens and securing the fifth position worldwide. Conversely, Kimi K2.5 experienced a slight decline, dropping to sixth place with 7.4 trillion tokens, though still recording an 8% increase from the prior week.
The data underscores a growing influence of Chinese AI models on the global stage, outperforming American models in recent weekly call volumes and establishing a new competitive landscape in AI development.
(For further details, please credit Kuaikeji. All illustrations are sourced from IFeng Images.)
ChatGPT
Perplexity
Gemini AI
Grok AI
