Blow to Hegemony: Huawei Open-Sources Training Code for Pangu-2.0 AI Model

While US labs hide their developments under the guise of "safety," China is striking a geopolitical blow through open-source. On September 28, 2026, Huawei shocked the IT community by publicly releasing the full code repositories for training its advanced `openPangu-2.0` model.

It's not just weights being published, but the very foundation of AI creation: algorithms for pretraining, Supervised Fine-Tuning (SFT), and Reinforcement Learning (RL). This is a masterful macroeconomic move by Beijing. By giving away the tools to train powerful LLMs for free, Huawei is luring the global community of independent developers into its architectural ecosystem, undermining the proprietary monopoly of OpenAI and Google. American sanctions have resulted in China becoming the de facto main sponsor of global open-source AI, shaping an alternative technological stack for countries unwilling to depend on Silicon Valley's clouds.

Source: Huawei