Accordingly, the division, Noah Ark Lab, issued a statement, dismissing allegations that a version of their major language model had copied elements from a Alibaba model last weekend, after an entities named HonestAGI posted an English article on the Github source code sharing platform, saying that Huawei's Pangu Pro Moe (Mixture of Experts) model had "unusual correlation" with Alibaba's Qwen 2.5 14B model. The report said that this shows that Huawei's model was created through "upcycling" and was not trained from scratch.
The above information has caused widespread discussion in the AI world and in Chinese media.
In the above statement, Noah Ark Lab said that this model "is not based on additional training from other manufacturer models". They affirmed that they have implemented important innovations in architectural design and technical features. At the same time, it said that this is the first large-scale model built entirely on Huawei's Ascend chip.
The department also said that its development team has strictly followed open source licensing requirements for any third-party source code used, but did not specify which open source models they have looked at.
It is known that the launch of the open source model R1 by startup company DeepSeek in January 2025 caused a stir in Silicon Valley with low costs and performance equivalent to other expensive models. On the other hand, DeepSeek has also fueled a fierce competition between Chinese technology giants for developing competitive-priced AI products.
Previously, one of the small models in Alibaba's Qwen 2.5 series - Qwen 2.5-14B released in May 2024 - could be deployed on both personal computers (PC) and smartphones.
While Qwen targets popular consumers and has services similar to ChatGPT, Huawei's Pangu models tend to be used more in government operations as well as the financial and manufacturing sectors.
Although Huawei soon entered the big language model market with the initial launch of Pangu in 2021, it was later said to have fallen behind its competitors. Huawei launched its Pangu Pro Moe models on the Chinese developer platform GitCode in late June. The goal is to promote Huawei's application of AI technology by providing developers with free access to these models.
