DeepSeek versus ChatGPT-4.0: A Comparative Analysis in Ophthalmology (Cross-Sectional Study) DeepSeek’e Karşı ChatGPT-4.0: Oftalmolojide Karşılaştırmalı Bir Analiz (Kesitsel Çalışma)
Turkiye Klinikleri Journal of Medical Sciences, cilt.46, sa.2, ss.164-169, 2026 (Scopus, TRDizin)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 46 Sayı: 2
- Basım Tarihi: 2026
- Doi Numarası: 10.5336/medsci.2025-115063
- Dergi Adı: Turkiye Klinikleri Journal of Medical Sciences
- Derginin Tarandığı İndeksler: Scopus, Central & Eastern European Academic Source (CEEAS), CINAHL, EMBASE, TR DİZİN (ULAKBİM), Biomedical Reference Collection: Corporate Edition (EBSCO), Health Research Premium Collection (ProQuest)
- Sayfa Sayıları: ss.164-169
- Anahtar Kelimeler: Artificial intelligence, large language models, medical education, ophthalmology
- Ankara Üniversitesi Adresli: Evet
Özet
Objective: To evaluate the performance of ChatGPT4.0 and DeepSeek in answering multiple-choice questions (MCQs) from the Review of Ophthalmology book. Material and Methods: In February 2025, 629 text-based MCQs were entered to ChatGPT-4.0 and DeepSeek. The evaluated subsections included optics (n=65), pharmacology (n=55), embryology/pathology (n=40), neuro-ophthalmology (n=64), pediatrics/strabismus (n=71), orbit/lids/adnexa (n=55), cornea/external disease (n=60), uveitis (n=55), glaucoma (n=45), anterior segment (n=67), and posterior segment (n=52). The percentage of correct responses was calculated per section, and an overall assessment score was determined. Results: ChatGPT-4.0 correctly answered 74.6% (469/629) of questions, while DeepSeek achieved 76.2% (479/629). No statistically significant difference was observed in the response accuracy rates between the artificial intelligence models (p=0.556). Both models demonstrated their lowest accuracy in optics, with ChatGPT-4 achieving 64.6% (42/65) and DeepSeek scoring 63.1% (41/65). In contrast, ChatGPT-4 exhibited its highest accuracy in pharmacology (87.3%, 48/55), while DeepSeek achieved its best performance in embryology/pathology (90.0%, 36/40). ChatGPT-4 and DeepSeek substantial agreement with the answer key, respectively (Cohen’s κ=0.660, Cohen’s κ=0.682). Conclusion: To our knowledge, this study is the first to compare ophthalmology-related MCQs from ChatGPT-4 and DeepSeek, 2 of the most popular AI chatbots. Both chatbots had an acceptable performance in responding to ophthalmology related MCQs. Therefore, it is reasonable that ophthalmologists suggest either chatbot as a source of supplementary education.