Who Knows Anatomy Best? A Comparative Study of ChatGPT-4o, DeepSeek, Gemini, and Claude.

This study evaluates the performance of ChatGPT-4o (OpenAI), DeepSeek-v3 (DeepSeek), Gemini 2.0 (Google DeepMind), and Claude 3.7 Sonnet (Anthropic) in answering anatomy questions from the Turkish Dental Specialty Admission Exam (DUS). The study aims to compare their accuracy, response times, and answer lengths. A total of 74 text-based multiple choice anatomy questions from the Turkish Dental Specialty Admission Exam (DUS) administered between 2012 and 2021 were analyzed in this study. The questions varied in difficulty and included both basic anatomical identification and clinically oriented scenarios, with a majority focusing on head and neck anatomy, followed by thorax, neuroanatomy, and musculoskeletal regions, which are particularly relevant to dental education. The accuracy of answers was evaluated against official sources, and response times and word counts were recorded. Statistical analyses, including the Kruskal-Wallis and Cochran's Q tests, were used to compare performance differences. ChatGPT-4o demonstrated the highest accuracy (98.6%), while the other models achieved the same rate of 89.2%. Gemini produced the fastest responses (mean: 4.47 s), whereas DeepSeek generated the shortest answers and Gemini the longest (p = 0.000). The differences in accuracy, response times, and word count were statistically significant (p < 0.05). ChatGPT-4o outperformed other models in accuracy for DUS anatomy questions, suggesting its superior potential as a tool for dental education. Future research should explore the integration of LLMs into structured learning programs.