Integrating advanced AI-driven question answering (QA) systems into educational settings offers significant potential for enhancing learning experiences. This study focuses on developing and optimizing an educational QA system using the T5-base model, a versatile text-to-text transformer known for its robust performance in natural language processing tasks. In addition to T5, other prominent large language models (LLMs) such as GPT-3, GPT-4 and BERT are also evaluated to compare several vital metrics comprehensively. By employing deep learning techniques such as Transformer architecture and sequence-to-sequence (Seq2Seq) models, the QA system is designed to provide contextually relevant and accurate responses to educational queries. The T5 model has been fine-tuned and optimized through experiments, incorporating specific hyperparameters—Batch Size, Epoch, Warmup Steps, and Weight Decay —to balance memory efficiency and response speed. The results indicate that, despite its smaller size compared to larger models like GPT-3, the T5-base model effectively generates meaningful answers, demonstrating its potential utility in educational applications. Compared to larger models like GPT-3, T5 provides a balanced solution, delivering accurate answers while minimizing resource consumption, thereby making it a practical choice for educational settings. The evaluation results emphasize the importance of selecting models that balance semantic similarity (Cosine Similarity) with deeper contextual understanding (BERT Score), which is crucial for educational contexts. This research evaluates the effectiveness of the T5-base model and provides a benchmark for assessing other LLMs in educational QA applications, highlighting the need for a balanced approach in model selection, considering factors such as performance, resource efficiency, and the specific requirements of educational environments. The evaluation results emphasize the need for a balanced approach in model selection, considering factors such as performance, resource efficiency, and the specific requirements of educational environments. This study contributes to creating more innovative and effective educational tools by enhancing the understanding of AI-driven QA systems in education.