Conceptual diagram of AI Photo: VCG

Chinese large language-models (LLMs) have continued to drive global artificial intelligence (AI) development, as the Global Times learned from OpenRouter, a popular AI gateway for developers, on Sunday that this week, on its LLM leaderboard, six out of the top 10 models are Chinese.

According to LLM Leaderboard, which compares the most popular models on OpenRouter, this week, six Chinese LLMs, including Step 3.5 Flash, MiniMax M2.5 and DeepSeek V3.2, were among the top 10 models.

The market share ranking, which compared OpenRouter token share by model author, showed that on Sunday, Xiaomi ranked first with a market share of 21.4 percent, followed by Google with 12 percent.

Also this week, China's open-source model Kimi from Chinese company Moonshot appeared in an overseas AI company's model, drawing attention from Elon Musk.

Code-generation start-up Cursor on Friday said in a post on X that its AI model Composer 2 is available. A netizen using the name "Fynn" cited the company's post, saying that he "was messing with the OpenAI base URL in Cursor and caught this: accounts/anysphere/models/kimi-k2p5-rl-0317-s515-fast," noting that "so composer 2 is just Kimi K2.5 with RL [reinforcement learning]."

Musk commented on Fynn's post, saying that "Yeah, it's Kimi 2.5." 

On Saturday, Cursor's founder Aman Sanger said on X that "We've evaluated a lot of base models on perplexity-based evals and Kimi k2.5 proved to be the strongest!" He added that "It was a miss to not mention the Kimi base in our blog from the start. We'll fix that for the next model."

Sanger's remarks came after the official account Kimi.ai posted on X on Saturday saying "Congrats to the @cursor_ai team on the launch of Composer 2!" It said that "Seeing our model integrated effectively through Cursor's continued pretraining & high-compute RL training is the open model ecosystem we love to support," noting that Cursor accesses Kimi-k2.5 "as part of an authorized commercial partnership."

An X user named Atlasis commented: "A Chinese open-source model quietly became the backbone of the hottest coding tool in the world and most people didn't even notice until the congratulations tweet. The real AI race isn't labs vs labs, it's who can build the best product on top of someone else's model."

Tian Feng, former dean of SenseTime's Intelligence Industry Research Institute, told the Global Times on Sunday that from the immense popularity of Chinese open-source models on international platforms to this weekend's Cursor-Kimi incident, it all demonstrates that Chinese LLMs are continuously driving global AI development through extreme performance with low inference costs.

"In addition, the advocacy by Chinese LLMs for open cooperation and ecosystem win-win is no longer merely a strategic slogan — it has been validated as an absolute advantage by the global market's vote," Tian noted.