Retrieval-Augmented Generation (RAG) is an emerging paradigm that enhances the performance of Large Language Models (LLMs) by integrating external knowledge retrieval into their generative processes. One of the key challenges in optimizing RAG systems is the selection of an appropriate retrieval dataset, which directly affects model accuracy, retrieval efficiency, and response coherence. The chunk size plays a fundamental role in the performance of RAG systems, as it directly impacts retrieval efficiency, response quality, and computational costs. The Neural Bridge RAG Dataset 12000 was used to test the effect of a chunk size in our experiment.