DeepSeek, a Chinese artificial intelligence (AI) company, has achieved the highest usage of its AI models globally.
According to data from OpenRouter, the total usage of large language models worldwide reached 188 trillion tokens from October 19 to 25, marking a 14.63% increase from the previous week, as reported by Chinese media outlet Caixin on October 26.
DeepSeek's 'V4.1-Flash-1017' topped the weekly usage chart with 33.6 trillion tokens, an increase of 801% compared to the prior week. All five top spots were occupied by Chinese AI models, with DeepSeek's 'V4.1 Flash' in second place at 18 trillion tokens. Minimax's free version 'M3.1 Flash' ranked third with 13.7 trillion tokens, followed by Zhiyu AI's 'GLM 5.3 Flash' at 13.2 trillion tokens in fourth, and Tencent's 'Hy4 Preview' at 8.97 trillion tokens in fifth.
Chinese AI models showed a significant lead in usage, with a total of 93.54 trillion tokens, an 18.1% increase from the previous week. In contrast, U.S. AI models recorded 22.47 trillion tokens, a 29.4% rise. The usage of Chinese models was approximately 4.2 times that of the U.S., marking 26 consecutive weeks of leading the U.S. Overall, Chinese models accounted for 49.8% of total usage, while U.S. models represented 11.9%.
This data reflects the token volume processed through OpenRouter, a platform that connects various AI models. The figures are based on input and output tokens generated by developers and businesses utilizing AI models via API. Therefore, it does not encompass all AI service usage globally, as models used directly without going through OpenRouter are not included.
Developers can compare performance and costs of different models on OpenRouter, selecting those suitable for coding tools or AI agents. The high usage of Chinese models is attributed to a combination of competitive pricing, performance in specific tasks, and the expansion of their use within the developer ecosystem.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.