Huawei Refutes Claims of Using Rival Data to Train Open-Source Pangu AI Model
Huawei denies Alibaba Cloud's claim that its Pangu Pro MoE 72B AI model was trained on Alibaba’s data. The model was developed using Huawei’s own Ascend chips and open-source code.

Huawei Responds to Allegations Over Pangu AI Model Training Data
Huawei recently open-sourced its Pangu AI models, sparking a debate after rival Alibaba Cloud accused the company of using its proprietary data for training. Specifically, Alibaba Cloud alleges that Huawei’s Pangu Pro MoE 72B hybrid expert model was trained on Alibaba’s Qianwen Qwen-2.5 14B model, raising copyright concerns.
Huawei firmly denies these claims. The company clarified that while the Pangu 72B model incorporates code from various open-source models, it was not trained on data from other companies. Instead, Huawei emphasized that the entire development and training process was carried out on its own Ascend chips.
Details on Huawei’s Pangu AI Models
On June 30, Huawei released two Pangu AI models under an open-source strategy. These models include the Pangu 7B dense model and the Pangu Pro MoE 72B hybrid expert model. This approach allows developers to access the original code and customize the models to fit their specific application or coding needs.
Huawei highlighted that the Pangu Pro MoE 72B model integrates advanced architectural designs and technical features. It was trained using Ascend GPUs and NPUs, which help address load balancing challenges in large-scale distributed training and improve overall training efficiency.
What This Means for Developers
- Access to open-source AI models with the flexibility to modify and use them in various applications.
- Confidence in Huawei’s claim that the models are built and trained independently on proprietary hardware.
- Potential for improved AI training performance thanks to hardware-optimized designs.
For developers interested in exploring AI models and training techniques, resources and courses on AI development can be found at Complete AI Training.
Huawei’s focus on hardware integration and open-source collaboration reflects a broader trend of tech companies pushing for transparent and accessible AI development tools, despite competitive tensions in the industry.