Drug discovery has advanced significantly with the emergence of Generative Artificial Intelligence (AI) technologies, outpacing traditional methods. Transformer-based generative models employ an attention mechanism to recognize contextual information in molecular data, which enhances their prediction accuracy. These models offer efficient and accurate solutions to complex molecular interaction problems, promising to accelerate drug discovery and improve global health outcomes. Moreover, they can act as powerful tools addressing various tasks in the drug discovery pipeline, including drug-target interaction prediction, molecular property prediction, molecular optimization, retro-synthesis prediction, and more. The current work thoroughly analyzes Transformer-based architectures for predicting molecular properties, focusing on datasets that use the Simplified Molecular Input Line Entry System (SMILES). Exhaustive experimentation has revealed the exceptional performance of Transformer Generative AI models in precisely capturing molecular features and interactions essential for accurate predictions. This study merges AI approaches with biochemical datasets to enhance drug property prediction and underscores the potential of Generative AI in transforming drug discovery processes.