← Founder Notes
Archive

The model usage map flipped and almost nobody is talking about it. china's llm call volume has now…

Yethikrishna ROriginal on Threads

the model usage map flipped and almost nobody is talking about it. china's llm call volume has now beaten the us for 20 straight weeks, and openrouter's top-10 is majority chinese models after being dominated a year ago.

the west still leads on frontier demos, the volume is elsewhere.

Context

Yicai Global reported on September 17, 2026 that global LLM call volume on OpenRouter reached 127 trillion tokens from September 7 to 13, with Chinese models accounting for nearly 61.2 trillion tokens against 21.8 trillion for US rivals, Chinese up 7.9 percent week on week and US models up 32 percent. Its headline says this is the 20th straight week Chinese models topped US rivals. A secondary blog from the same date says seven of the top 10 models that week were Chinese.

How it compares

The measure is tokens processed through OpenRouter in one week and not call counts. Chinese and US models together are about 83 trillion of the 127 trillion, so roughly 44 trillion belong to other models. The 20 week streak is the outlet's headline and the underlying weekly series was not read. The ranking basis for the top 10 was not stated, and nothing was found on a year-ago top 10. OpenRouter is one routing platform with its own user mix and prices, so this is not a general adoption conclusion. The same blog notes the top five on an Artificial Analysis index are still Western, so volume does not show capability.

Watch next

  • OpenRouter's own rankings page for the weekly series, and a year-ago comparison.

Sources

  1. Yicai Global: Chinese LLMs top US rivals in call volume on OpenRouter for the 20th straight week (September 17, 2026)www2.yicaiglobal.com
  2. redeep: China AI inference usage vs US (September 17, 2026)redeep.blog

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 21 September 2026 at 22:47 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/the-model-usage-map-flipped-and-almost-nobody-DdjrT4TgoQ7" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="The model usage map flipped and almost nobody is talking about it. china's llm call volume has now…"></iframe>

More notes