One-sentence summary
At the Wuzhen Summit: the top five weekly LLM models by calls were all Chinese, totaling 93.54 trillion tokens versus 22.47 trillion for US models, extending a 26-week lead in total usage.
The Scope
Using OpenRouter's monitoring, the 2026 Wuzhen Summit disclosed this week's China-US comparison of LLM usage.
Core Comparison
| Metric | Chinese models | US models |
|---|---|---|
| Weekly usage total | 93.54T tokens | 22.47T tokens |
| Global top-five seats | All five are Chinese | None in top five |
| Consecutive lead | Outpaced US for 26 weeks running | — |
Key Differences
Chinese models win developer preference at lower inference cost; budget-limited overseas teams and independent developers widen the gap, while US models hold no top-five seat.
Sources
- Guangming Online: Wuzhen voices open source
- CNR: Wuzhen voices open source
