According to a May 6 report on the website of the Japanese Economic News, Qwen, the artificial intelligence (AI) large model developed by AliCloud under Alibaba Group, is gaining increasing influence in Japan. The Qwen model, which is open-sourced to the outside world, is being adopted by Japanese startups such as ABEJA. In the "AI Model Rating" released by the Japanese Economic News in April, Qwen ranked sixth. The report noted that on April 17, ABEJA released a new reasoning model based on the Qwen foundation model. In the comprehensive ranking published by Nikkei Digital Governance, a professional media under the Nikkei News Agency, testers scored the performance of major AI models from countries including the U.S., China, and Japan using Japanese language. Qwen2.5-Max, released in January this year, ranked sixth among all 113 models. It performed excellently not only in basic language skills such as grammar but also achieved outstanding results in logical reasoning and mathematics. Open-source models can be used by external companies for training and developing their own AI models. The highest-scoring model (ranked 21st overall) among the participating Japanese enterprises, ABEJA-Qwen2.5-32B, was also developed based on Qwen. Among Japan's startups, ELYZA, originating from the University of Tokyo, released a model based on Qwen on May 1. Lightblue, also originating from the University of Tokyo, and Axcxept, established in Sapporo, are also using the Qwen model. Li Zhihui, head of Nomura Research Institute's China Innovation Research Program, pointed out that Qwen "can achieve high-precision output with a small dataset, thus receiving high evaluations in multiple indicators for Japanese language capabilities." Qwen has also drawn attention from more countries. According to AliCloud Company, over 100,000 derivative models of Qwen have been developed on the AI model sharing website Hugging Face. Li Zhihui believes that "the performance of Qwen has been recognized worldwide compared to other Chinese open-source models." Masayu Yokosawa, head of AliCloud's Japanese branch, said, "We hope to introduce successful experiences from China." The company has been in negotiations with several Japanese enterprises. According to Yokosawa, they hope the number of Qwen model connections in Japan will exceed one thousand within three years. On April 29, AliCloud released its latest model Qwen3. The pre-trained data volume is twice that of Qwen2.5. At the same time, supporting 119 languages and dialects demonstrates the rapid progress of the Qwen model. It was reported that another Chinese startup, DeepSeek, is also rising in the open-source market. Its DeepSeek model is being widely integrated into industries such as government, automobiles, and finance. The advantage of open-source models lies in having a large number of developer users, making them easy to validate and improve. The trend of Chinese AI rapidly catching up with OpenAI in terms of performance through an open strategy will continue. (Translated/Compiled by Liu Lin) [Image: https://p3-sign.toutiaoimg.com/tos-cn-i-axegupay5k/10be51449cb04c05a55b493604c19397~tplv-tt-origin-web:gif.jpeg?_iz=58558&from=article.pc_detail&lk3s=953192f4&x-expires=1747194968&x-signature=YgCk1gxLEE1U8ybG5leshvbXgD4%3D] On March 28, during the annual session of the 2025 Zhongguancun Forum held at the permanent exhibition hall in the Zhongguancun Exhibition Center, visitors viewed the "Artificial Intelligence+" display area. (Xinhua) Original article: https://www.toutiao.com/article/7501517886469554688/ Disclaimer: This article solely represents the author's personal views. Please express your opinions by clicking the "Like/Dislike" buttons below.