下载超10亿、衍生模型破20万,阿里千问刷新全球开源模型纪录
Bei Ke Cai Jing·2026-01-21 05:10

Core Insights - Alibaba's Qianwen derivative models have surpassed 200,000, making it the first open-source large model to achieve this milestone globally [1] - The download count for the Qianwen series models has exceeded 1 billion, averaging 1.1 million downloads per day, significantly surpassing the American Llama model [1] Model Impact and Development - The number of derivative models is considered a key indicator of the influence of open-source models [1] - Since 2023, Alibaba has open-sourced nearly 400 large models, leading to a surge in new models developed by global developers based on the Qianwen model [1] - These new models are further open-sourced, stimulating the emergence of new technologies and applications [1] Community Engagement - The Stanford team led by Fei-Fei Li has trained a new model, s1, based on Qianwen, expanding the exploration of AI inference within the open-source community [1] - Developers are adding over 200 new derivative models daily based on Qianwen, covering various fields such as robotic control, code generation, comic post-production, and multilingual translation [1]