Artificial Intelligence (AI) technology based on Large Language Models (LLMs) in education offers significant potential to overcome the limitations of conventional learning, thereby enhancing personalized and adaptive learning experiences for students. The purpose of this research is to develop and evaluate the feasibility of a Chatbot. LLMs-based AI education specifically designed to support the Q&A activities of Science learning (Physics, Chemistry, and Biology) at the Senior High School (SMA) level. This AI chatbot was developed as a learning aid that can provide accurate answers, contextual explanations, and relevant examples in real-time to student questions at SMA Madania Bogor. The type of research used is Research and Development (R&D). The development model used specifically for software is Waterfall with the following stages: Requirements analysis, system design, implementation, testing, deployment, and maintenance. Data collection techniques include observation, interviews, questionnaires, and documentation. The method for developing media in this study is by designing a Chatbot architecture that has previously been analyzed data, selecting and fine-tuning LLMs that are suitable for the Science domain, and implementing an intuitive user interface, then the media is tested for accuracy and feasibility by expert review experts in the field of IT and Design and Chatbots that have been installed on a server or platform with continuous monitoring from the developer for program maintenance . The results of this study found that the AI Chatbot device based on Large Language Models (LLMs) is effective in improving students' question and answer skills in science learning activities at SMA Madania Bogor.