Huawei's AI research division, Noah's Ark Lab, has refuted accusations that its Pangu Pro large language model (LLM) copied elements from Alibaba's Qwen 2.5 14B model. The lab asserted that Pangu Pro was independently developed and trained, highlighting key innovations in its architecture and technical features.
The denial follows claims of an "extraordinary correlation" between Huawei's Pangu Pro Moe (Mixture of Experts) model and Alibaba's Qwen 2.5 14B, as stated in a paper posted on GitHub. This report suggested potential copyright violation and questioned Huawei's investment in training the model. Huawei countered that Pangu Pro was not based on incremental training of other manufacturers' models and was designed for the Ascend hardware platform.
Pangu Pro models are primarily used in government and finance, while Alibaba's Qwen models are more consumer-focused, such as chatbot services. Huawei open-sourced its Pangu Pro Moe models on GitCode in late June to boost AI tech adoption.
