Huawei’s artificial intelligence research arm, Noah Ark Lab, has denied allegations that its Pangu Pro Moe large language model (LLM) copied elements from Alibaba’s Qwen 2.5 14B, asserting that the model was independently developed and trained.
The denial came in a statement issued Saturday, following the release of a paper by an entity named HonestAGI on GitHub a day earlier. The paper claimed that Huawei’s Pangu Pro Moe model showed “extraordinary correlation” with Alibaba’s Qwen, suggesting the model was derived through “upcycling” rather than being trained from scratch.
The paper’s authors also alleged potential copyright violations, fabrication of details in technical reports, and misleading claims about Huawei’s investment in training the model — claims that quickly drew attention across AI circles and Chinese tech media.
In response, Noah Ark Lab maintained that Pangu Pro Moe was “not based on incremental training of other manufacturers’ models”, and that it featured “key innovations in architecture design and technical features.” The lab also noted that the Pangu Pro Moe is the first large-scale model fully built on Huawei’s Ascend chips, showcasing its in-house hardware and software integration.
While acknowledging the use of open-source components, Huawei emphasized that its development team had complied with all applicable license terms, though it did not specify which open-source resources were referenced.
Alibaba has not commented on the matter, and Reuters was unable to reach HonestAGI or verify its affiliations.
Huawei initially launched the Pangu series of models in 2021, making it one of China’s early entrants into the LLM space. However, it has since been viewed as lagging behind competitors. In June 2025, Huawei open-sourced the Pangu Pro Moe models on the Chinese developer platform GitCode to encourage adoption among developers.
Unlike Alibaba’s Qwen, which powers consumer-facing chatbot services similar to ChatGPT, Huawei’s Pangu models are primarily aimed at applications in government, finance, and manufacturing sectors.
The incident highlights the competitive intensity in China’s AI sector, which has accelerated since the release of Chinese startup DeepSeek’s low-cost open-source model earlier this year, triggering a wave of innovation and scrutiny among the country’s tech giants.