Huawei AI Lab Denies Copying Alibaba Qwen Model
Pangu is a large learning language model featuring 1.085 trillion parameters, developed by Huawei.It was first announced during the Huawei Developer Conference on July 7, 2023. The model was developed within Huawei Mindspore 5 - an open-source deep learning training interface.
Alibaba's Qwen on the other hand, is a family of large language model developed by Alibaba Cloud in 2023.
On July 4, 2025, Pangu Pro MoE (Mixture of Experts) model was alleged to be similar with Alibaba Qwen- 2.5 model in a GitHub post by an unidentified member of HonestAGI (a newly created group). This claim was based on "model fingerprinting" technology.
According to the post, it claimed that Huawei's Pangu Pro MoE (Mixture of Experts) model shows Extraordinary correlation (0.927) with Qwen- 2.5 14B. It suggests that that the Pangu Pro MoE (Mixture of Experts) was derived through upcycling rather than training from the scratch.
However, Huawei Noah's Ark research lab, the team behind Pangu, has denied these allegations
Comments