A method called dynamic multi-group and multi-level based on ontology was proposed to build user model according to national crop germplasm resourcesinfrastructure. A user node was first created, and an initial multi-group user model was built on the basis of user registration information and domain ontology. Then site log was mined to extend and update the user model regularly. Finally synonyms extension and different languages semantic expansion was implemented using common ontology WordNet, so as to further optimize the user model. The core domain ontology of national crop germplasm resources infrastructure was built under the environment of open source software protégé. User information was added to the ontology and formed the initial user model. User interest degree to resources was calculated, and time factor was added to the user model to indicate the interest change of users, so as to dynamically update the user model. The user model of dynamic multi-group and multi-level based on ontology could provide personalized services to different users, and could help improve service quality of national crop germplasm resources infrastructure.