Comparative Performance of DeepSeek and ChatGPT-4o in the Chinese Nursing Licensing Exam

Background: The rapid advancement of open-source models such as DeepSeek presents significant potential for nursing education, though its effectiveness remains unexplored. Method: This study evaluated the performance of DeepSeek-R1, DeepSeek-V3, and ChatGPT-4o using 720 questions from three mock exams for the 2025 Chinese Nursing Licensing Exam. The analysis focused on answer accuracy and logical consistency. Results: DeepSeek models demonstrated better performance than ChatGPT-4o. Deep-Seek-R1 achieved an accuracy range of 87.5% to 95.8%, and DeepSeek-V3 achieved an accuracy range of 85.8% to 94.2%, significantly outperforming ChatGPT-4o's range of 67.5% to 78.3%. Moreover, DeepSeek-R1 (83.1% to 95.9%) and DeepSeek-V3 (84.6% to 96.8%) exhibited higher accuracy across different question types and provided more logically consistent, professionally grounded explanations than ChatGPT-4o (61.5% to 75.9%). Conclusion: The findings highlight the substantial potential of DeepSeek as a valuable educational tool for nursing and other disciplines. However, its limitations should be carefully considered, necessitating further exploration to fully understand its applications.