(Yicai) March 23 -- Chinese large language models are dominating global usage of artificial intelligence, staying ahead of US competitors for the third week in a row, according to rankings released by OpenRouter.

Among the top nine models by global usage volume last week, five were Chinese, OpenRouter’s leaderboard showed today. The top four were Xiaomi Group’s MiMo-V2-Pro, Step 3.5 Flash from Shanghai Jieyue Xingchen Intelligent Technology, M2.5 from MiniMax, and DeepSeek’s V3.2. Z.ai’s GLM5 Turbo took the sixth spot.

Total usage was 7.359 trillion tokens, up 57 percent from 4.69T in the previous week. 

The US models in the top nine were Claude Sonnet 4.6 and Claude Opus 4.6 from Anthropic, along with Gemini 3 Flash Preview and Gemini 2.5 Flash from Google, with total usage of 3.536T, up 7.3 percent from 3.294T in the prior week. 

China’s open-source models saw the highest global download volume last year. They have proved popular partly through being free to use, but also because they allow developers worldwide to fine-tune them. 

This has significantly lowered the barriers to using AI globally and made the technology truly a public good, Joseph Tsai, chairman of Chinese tech giant Alibaba Group Holding, said at the China Development Forum’s annual conference. 

There was further evidence of the global influence of Chinese models last week, after US-based AI coding firm Cursor was found to have based its latest release on Kimi 2.5, a model released earlier this year by Chinese startup Moonshot AI. 

Users soon spotted the similarities, with one posting online that Cursor’s Composer 2 is just Kimi 2.5 with reinforcement learning. This drew the attention of Tesla boss Elon Musk who commented on the post, saying "Yeah, it's Kimi 2.5."

Cursor swiftly admitted the connection. "It was a miss to not mention the Kimi base in our blog from the start. We'll fix that for the next model," founder Adam Sanger said in a post on X.

Editor: Tom Litting