This paper investigates the efficacy of transformer-based models—BERT, T5, and GPT-2—in text summarization and fine-grained sentiment analysis. Employing these models, we explore their performance in abstractive text summarization, with each model presenting distinct approaches. BERT, leveraging bidirectional context understanding, excels in extracting pivotal information, while T5 adopts a text-to-text translation approach, generating coherent summaries. GPT-2, with its autoregressive architecture, predicts sequential words to produce cohesive summaries. In fine-grained sentiment analysis, these models decode intricate sentiments within texts. BERT utilizes contextual embeddings to discern nuanced relationships between words, T5 fine-tunes for sentiment analysis efficiently, and GPT-2 adapts to text classification challenges for sentiment interpretation. Our study highlights the capabilities and diverse methodologies of these models in capturing contextual understanding and nuanced sentiments. The findings underscore BERT’s contextual grasp, T5’s versatile text-to-text framework, and GPT-2’s adaptability for sentiment analysis. These insights pave the way for leveraging transformer models in enhancing text understanding and sentiment interpretation in various applications.