Shvartz et al. compare how well GPT-4 and GPT-4o perform on ophthalmology board exam questions in both English and French, especially in interpreting images and different question types across subspecialties. Shvartz et al. provided GPT-4 and GPT-4o a set of English and French ophthalmology exam questions from the Israeli state certification-level board examinations to answer. Shvartz et al. then compared how accurate each model was by comparing its performance with that of residents from 2021-2023 on the same examination. Shvartz et al. found that GPT-4o outperformed GPT-4 in both English and French on the ophthalmology board exam questions. Higher performance was significantly evident in image-based questions and across various subspecialties. However, both models still had limitations in accuracy, highlighting that further improvements are needed for more efficient clinical application. Overall, Shvartz et al. highlight that while GPT-4o shows improved accuracy over GPT-4 in multilingual ophthalmology exams, caution is still needed due to possible errors in clinical applications.