Generative AI models are transforming biomedical research by enabling thorough analysis of medical literature to evaluate the relevance of medications. This study compares three sophisticated models: BioGPT, GPT-4, and PubMed BERT, in terms of their understanding of medication names and their recognition of applications. Each model is evaluated based on accuracy, contextual relevance, and flexibility using a curated dataset that includes annotated medication names and their corresponding uses. BioGPT, designed specifically for biomedical tasks, excels with specialized terminology but struggles in ambiguous contexts. GPT-4, as a general-purpose model, demonstrates remarkable adaptability and language understanding but lacks deep specialization in biomedical domains. PubMed BERT, tailored to PubMed literature, achieves a compromise between accuracy in domain-specific contexts and contextual awareness, benefiting from extensive training on biomedical texts. This analysis highlights the trade-offs between general versatility and specialized expertise, providing insights into choosing suitable models for various healthcare and pharmaceutical applications. These findings guide the integration of generative AI into clinical decision-making and medical research activities.