Performance of ChatGPT-4, DeepSeek-R1, and Gemini-2.0 in undergraduate dental implantology education: a comparative study

Existing guidance on the application of artificial intelligence in implantology education was mostly based on clinical cases or scenarios, making them more suitable for postgraduate or continuing education. To date, no study has evaluated the capability and quality of large language models’ (LLMs) responses to foundational implantology education, considering both core textbook knowledge and open-ended literature research. This study aims to compare the performance of different LLMs in answering professional implantology questions. Fifty multiple-choice questions (MCQs) including four knowledge sections and 10 open-ended questions (OEQs) for undergraduate dental implantology education were prepared. Each question was presented to three models (including ChatGPT-4.0, DeepSeek-R1, and Gemini-2.0) three times for three consecutive weeks. MCQ responses were evaluated by calculating accuracy rates. OEQs responses were blindly evaluated for accuracy, completeness, and relevance based on a three-point Likert scale. Generalized Estimating Equations, Kruskal-Wallis H test, and Bonferroni-corrected pairwise comparison were used for statistical analysis (α = 0.05). LLM type significantly influenced performance on both MCQs and OEQs ( p < 0.05). For MCQs, the overall accuracy rate ranked as: ChatGPT-4 (84.67%) > DeepSeek-R1 (83.78%) > Gemini-2.0 (78.44%). For OEQs, the total score ranked as: Gemini-2.0 (1.63 ± 0.29) > DeepSeek-R1 (1.57 ± 0.36) > ChatGPT-4 (1.53 ± 0.30). Significant differences were observed in accuracy across different knowledge sections within the same LLM ( p < 0.001). The testing time only significantly affected OEQs, with Gemini-2.0 exhibiting a consistent weekly score increase ( p < 0.001). In MCQs, ChatGPT-4 outperformed both DeepSeek-R1 and Gemini-2.0; however, the opposite trend was observed for OEQs. Nevertheless, none of the three models demonstrated sufficient reliability for independent use in professional implantology education within the present study framework, highlighting the necessity of expert oversight. Developing domain-specific AI programs for dental implantology may help improve accuracy and consistency.

Authors

Institutions

Publication Details

Journal
BMC Oral Health
Published
2026-11-13
DOI
https://doi.org/10.1186/s12903-026-10357-8
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Performance of ChatGPT-4, DeepSeek-R1, and Gemini-2.0 in undergraduate dental implantology education: a comparative study

Jiakang Zhu, Jian Wei, Caini Wei, Zhixing Wu et al.
7 links
BMC Oral Health
Artificial Intelligence in Healthcare and Education
article

Performance of ChatGPT-4, DeepSeek-R1, and Gemini-2.0 in undergraduate dental implantology education: a comparative study

Jiakang Zhu, Jian Wei, Caini Wei, Zhixing Wu, Cui Huang
article en

Abstract

Existing guidance on the application of artificial intelligence in implantology education was mostly based on clinical cases or scenarios, making them more suitable for postgraduate or continuing education. To date, no study has evaluated the capability and quality of large language models’ (LLMs) responses to foundational implantology education, considering both core textbook knowledge and open-ended literature research. This study aims to compare the performance of different LLMs in answering professional implantology questions. Fifty multiple-choice questions (MCQs) including four knowledge sections and 10 open-ended questions (OEQs) for undergraduate dental implantology education were prepared. Each question was presented to three models (including ChatGPT-4.0, DeepSeek-R1, and Gemini-2.0) three times for three consecutive weeks. MCQ responses were evaluated by calculating accuracy rates. OEQs responses were blindly evaluated for accuracy, completeness, and relevance based on a three-point Likert scale. Generalized Estimating Equations, Kruskal-Wallis H test, and Bonferroni-corrected pairwise comparison were used for statistical analysis (α = 0.05). LLM type significantly influenced performance on both MCQs and OEQs ( p < 0.05). For MCQs, the overall accuracy rate ranked as: ChatGPT-4 (84.67%) > DeepSeek-R1 (83.78%) > Gemini-2.0 (78.44%). For OEQs, the total score ranked as: Gemini-2.0 (1.63 ± 0.29) > DeepSeek-R1 (1.57 ± 0.36) > ChatGPT-4 (1.53 ± 0.30). Significant differences were observed in accuracy across different knowledge sections within the same LLM ( p < 0.001). The testing time only significantly affected OEQs, with Gemini-2.0 exhibiting a consistent weekly score increase ( p < 0.001). In MCQs, ChatGPT-4 outperformed both DeepSeek-R1 and Gemini-2.0; however, the opposite trend was observed for OEQs. Nevertheless, none of the three models demonstrated sufficient reliability for independent use in professional implantology education within the present study framework, highlighting the necessity of expert oversight. Developing domain-specific AI programs for dental implantology may help improve accuracy and consistency.

BMC Oral Health
Wuhan University (CN), Tongji Hospital (CN), Huazhong University of Science and Technology (CN)
Science and Technology Program of Hubei Province
Quality education, Good health and well-being
Artificial Intelligence in Healthcare and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon