The project team responsible for the AI training platform, hailing from the Shenzhen Futian Higher Education Mega Center, worked in tandem with several partners to accomplish the full-parameter post-training of the DeepSeek-V4-Pro—a large-scale model boasting 1.6 trillion parameters. This milestone was achieved utilizing the domestically-produced AI computing power cluster, Ascend 910C. This accomplishment underscores the capability of domestic AI chips to facilitate the training of ultra-large parameter models on a global scale. It signals that domestic AI computing power is now poised to undertake large model training tasks of the highest caliber, with the relevant technical approach having been validated as feasible.