For stock trading, rely on Jin Qilin analyst reports — authoritative, professional, timely, and comprehensive, helping you uncover potential thematic opportunities! Chinese AI large models have led global token usage for 22 consecutive weeks: the anonymous model "Space Bunny" has surged into the top three, with tests suggesting it may come from MiniMax; Xiaomi MiMo-V2.5 and Tencent Hy3 have dropped off the rankings.

From September 21 to September 27, China's weekly large model token usage exceeded that of the United States for the 22nd consecutive week. Among the global top five by token usage, three are Chinese AI large models: DeepSeek V4.1 Flash, Zhipu GLM 5.3 Flash, and Tencent Hy4 preview. The anonymous model Space Bunny Alpha (太空兔) has surged onto the list, taking third place. Some tests suggest the model may originate from MiniMax.

Based on the latest OpenRouter data, the total global AI large model token usage last week (September 21 to September 27) reached 146 trillion tokens, up 13.18% week-over-week. Among the ranked AI large models, Chinese AI large models recorded weekly token usage of 62.22 trillion tokens, down 7.77% week-over-week; over the same period, U.S. AI large models recorded weekly token usage of 14.2 trillion tokens, down 0.07% week-over-week. China's weekly large model token usage has exceeded that of the United States for 22 consecutive weeks, firmly holding the top spot globally.

It was noted that last week, among the global top five by token usage, three were Chinese AI large models. Among them, DeepSeek V4.1 Flash held the top position for the second consecutive week, with weekly token usage reaching 19.6 trillion tokens, up 24% week-over-week; Zhipu GLM 5.3 Flash held second place for the second consecutive week, with weekly token usage reaching 16.3 trillion tokens, up 16% week-over-week; Tencent Hy4 preview ranked fourth, with weekly token usage reaching 9.64 trillion tokens, down 23% week-over-week.

In addition, the anonymous model Space Bunny Alpha (太空兔) surged onto the list, taking third place with weekly token usage of 13.9 trillion tokens. Space Bunny Alpha launched on OpenRouter on September 23. According to its introduction, the model focuses on fast inference and programming, features a 1-million-token context window, supports text, image, and video inputs, and offers adjustable reasoning intensity and tool-calling capabilities. Third-party tokenizer fingerprint testing showed that Space Bunny Alpha's token counts across 50 test strings were completely identical to the MiniMax series, leading the test to conclude that the model may originate from MiniMax. As of press time, the model's identity had not been officially confirmed.

Furthermore, Xiaomi MiMo-V2.6-Flash entered the rankings at eighth place, with weekly token usage reaching 5.51 trillion tokens. On September 22, Xiaomi officially released and open-sourced the MiMo-V2.6 series, comprising two native full-modality models: Pro and Flash. Among them, Pro has 1.02T total parameters and 42B activated parameters, while Flash has 309B total parameters and 15B activated parameters. Both models support a 1-million-token context and cover text, image, video, and audio inputs. According to Xiaomi, MiMo-V2.6-Pro scored 46 on the Artificial Analysis Intelligence Index (AA Comprehensive Intelligence Index), surpassing Kimi K3 and Qwen3.8 Max to become the strongest open-source model currently available; however, it still lags behind the strongest closed-source models, Claude Fable 5.1 and GPT-6 Astra. In addition, the MiMo-V2.6 series continues the API pricing of the V2.5 series. Among them, MiMo-V2.6-Pro, at comparable intelligence levels, is priced at only 1/20 to 1/60 of overseas models.

It is worth noting that Xiaomi MiMo-V2.5, which ranked sixth the previous week, and Tencent Hy3, which ranked seventh, both dropped off the list.

Disclaimer: The content and data in this article are for reference only and do not constitute investment advice. Please verify before use. Operate at your own risk. Cover image source: Meijing Media Library. Sina statement: This message is reposted from a Sina partner media outlet. Sina publishes this article for the purpose of conveying more information and does not imply endorsement of its views or confirmation of its descriptions. The article content is for reference only and does not constitute investment advice. Investors who act on it do so at their own risk.