Ali Tun Yi asked 3 open source downloads per month exceeding 12.5 million leading the global AI open source model.

The A.I.M. 3 (Qwen 3) large-scale, A.A.A.P. under the Ali Baba flag, has only one month to start from the source, and the global download has exceeded 12.5 million times, making it the most popular open-source model in the recent past. On mainstream AI platforms such as Hugging Face, the Magic Community and Ollama, four parameter size models under the three flags were downloaded more than a million times.

The Hugging Face data show that the four parameter size models for question 3 (0.6B, 8B, 30B, 32B) were favoured by developers, each with more than 1 million downloads. Among them, Qwen-32B ranked first on the Hugging Face ranking with a combined score of 43.02, surpassing many international competitions.

In addition, the number of multi-question series derivative models has reached 130,000, well above other open-source models, demonstrating their extensive application and potential for secondary development in the developers ‘ communities. In the magic community and the Ollama platform, question 3 was equally excellent, especially among Asian developers.

Question 3 was developed by the Ali Babayun Smart Group to cover large language models (LLM) and multi-modular models (MLLM) to support a wide range of data processing capabilities, such as text, image and voice. Its parameter sizes of 0.6B to 32B are designed to meet diversified needs from lightweight equipment to high performance servers.

The open-source version of thousands of questions 3 includes model weights, codes and data sets that support developers in customizing development based on them. Compared to other open-source models, Question 3 has significant advantages in terms of multilingual support (especially Chinese processing) and computational efficiency, and is appropriate for the context of globalization and localization.

The open-source heat of Pan-Yuji 3 provides Aliyun with a solid foundation for the global AI market. In the future, Ali plans to further expand the functionality of the ask-and-answer series by introducing models that support more modulations (e.g., video generation) and higher parameters.