Systematic Evaluation of ChatGPT-3.5 and ChatGPT-4.0 for Accuracy and Reliability in Veterinary and Animal Science Education

Large Language Models (LLMs) such as ChatGPT are increasingly used in veterinary and animal sciences education, yet systematic performance evaluations across academic domains remain limited. This study quantitatively compared ChatGPT-3.5 and ChatGPT-4.0 in terms of answer accuracy across Animal Sciences, Paraclinical Sciences and Clinical Sciences. A dataset comprising domain-specific questions was presented to both models and responses were scored on a 10-point accuracy scale by subject matter experts. Statistical analyses including mean scores, standard deviations, paired t-tests and one-way and two-way ANOVAs were applied to assess performance differences. Results showed that ChatGPT-4.0 consistently outperformed ChatGPT-3.5 across all domains, with mean differences of +2.0 in Animal Sciences, +2.0 in Paraclinical Sciences and +1.9 in Clinical Sciences, all statistically significant (p<0.001). Moreover, ChatGPT-4.0 exhibited lower variability, indicating more stable accuracy. The overall two-way ANOVA revealed significant main effects for both model version (p<0.001) and academic domain (p = 0.002), with no significant interaction effect. These findings demonstrate that ChatGPT-4.0 provides superior and more consistent performance in veterinary education question-answering compared to ChatGPT-3.5, supporting its adoption as a supplementary learning tool in domain-specific curricula.