On July 21st, Shanghai AI Laboratory proudly introduced the DeepLink technical solution for ultra-large-scale cross-domain mixed training. This groundbreaking solution facilitates long-term and stable mixed training of massive models, boasting hundreds of billions of parameters, across multiple intelligent computing centers spanning thousands of kilometers. For instance, it enables the seamless connection of intelligent computing centers located as far as 1,500 kilometers apart, such as those in Shanghai and Jinan, while also realizing interconnected intelligent computing centers and mixed training of large models in diverse locations including Beijing, Shanghai, and Guizhou. Presently, the DeepLink open computing system has been seamlessly integrated into the intelligent computing platforms of China Unicom, China Telecom, SenseTime, and Yidian, achieving robust operation within the "1 platform + N chips + X regions" model.