In this paper, we explore and compare the performance of three conversational chatbots based on large language models, focusing on solutions that are entirely open-source and accessible. The models were tested under realistic computational constraints using Google Colab’s free environment. The evaluation covered two tasks: closed-domain question answering (based solely on internal model knowledge) and standard question answering (with external context provided). Both technical performance and user feedback were considered. The technical assessment included over 250 questions, measuring accuracy, speed, informativeness, and language quality. A user survey involving some participants supplemented these findings.