Scientific and efficient biological breeding is of great significance for food security and economic benefits. For a long time, the complexity and diversity of crop breeding knowledge, the lack of unified management of breeding information, and the scattered distribution of data knowledge have brought challenges to the popularization and learning of breeding knowledge. In order to improve the level of breeding knowledge management, we constructed a maize breeding knowledge question and answer system that integrates fine-tuning large-scale model and Ritrival-Augmented Generation (RAG) technology. The system can be applied to intelligent breeding vehicle and directly question and answer the knowledge related to maize breeding, so as to solve the problem of difficult query of relevant knowledge in farmers’ breeding. In this process, we construct a question and answer data set and fine-tune the model, and integrate the inherent ability of the large model after fine-tuning and the knowledge retrieval enhancement technology to improve the response effect of the system. In the RAG process, we first chunk and vectorize the knowledge text and store it in the database, then integrate sparse retrieval and vector retrieval to enhance the recall ability, and finally combine the problem and the knowledge input model of retrieval to obtain professional answers about maize breeding. After the completion of the system construction, we conducted an expert evaluation of the response effect of the system. The results showed that the system could respond professionally and effectively to maize breeding knowledge.