Global AI model usage saw a slight decline last week, while competition among leading models continued to intensify.
According to data from OpenRouter, global AI model usage totaled 56.8 trillion tokens from July 27 to Aug. 2, down 2.07% week on week.
Chinese AI models recorded 28.13 trillion tokens in weekly usage, while U.S. models reached 4.38 trillion tokens. Chinese models have ranked ahead of U.S. models in weekly usage for 14 consecutive weeks.
The top five models by usage last week were all from China:
🥇 DeepSeek-V4-Flash — 7.22 trillion tokens
🥈 Xiaomi MiMo-V2.5 — 6.3 trillion tokens
🥉 Tencent Hy3 — 4.82 trillion tokens
4️⃣ DeepSeek-V4-Pro — 3.28 trillion tokens
5️⃣ Zhipu GLM-5.2 — 2.89 trillion tokens
OpenAI’s GPT-5.6 Luna appeared on the list for the first time, ranking eighth with 1.95 trillion tokens.
Meanwhile, pricing and cost efficiency are becoming a key part of AI competition. Chinese open-source models, new releases such as Kimi K3, and moves by global tech companies are adding pressure to an increasingly competitive market.
Which AI model are you using most often?
