CITIC Construction Investment has reported that DeepSeek has unveiled its latest flagship model, DeepSeek-V4-Pro. This new model boasts significant improvements in complex reasoning, programming, and agent capabilities. In contrast to the V4-Flash model, which prioritizes cost-effectiveness and efficiency, the V4-Pro is designed to tackle highly challenging tasks and agent scenarios with greater proficiency. It achieved a score of 53 on the Artificial Analysis Intelligence Index, surpassing the V4-Flash's score of 40. On August 14th, Zhipu also introduced its new-generation open-source flagship model, GLM-5.3, which continues to emphasize enhancements in programming, long-term task execution, and agent capabilities. Lately, domestic models like DeepSeek and Zhipu have been undergoing continuous iterations, rapidly closing the gap with overseas leading models in terms of reasoning, programming, and agent capabilities. The model ecosystem and commercialization process are gaining momentum, and it is anticipated that domestic models and domestic computing power will collaborate and develop in a mutually beneficial (synergistic) manner.