Medeniyet University Evaluates AI Vision-Language Models Against Medical Students
A comparative study conducted by the Department of Dermatology and Venereology at Istanbul Medeniyet University has evaluated the performance of state-of-the-art vision-language models (VLMs)—including ChatGPT-4o, ChatGPT-5, Gemini 2.5 Flash, and Gemini 3 Pro—against 252 fifth-year medical students. The findings, noted by Nature, reveal that while VLMs significantly outperformed human medical students on text-only examinations, AI performance was highly heterogeneous on image-based tasks. The study concluded that medical students demonstrated greater cross-modal consistency, suggesting that while AI is highly proficient in text-based assessments, its multimodal diagnostic competence remains uneven.
Source: Nature (via ResearchGate)