In order to identify military named entities a multi-models method was proposed. After detailed research and analysis on military text efficient military named entity recognition system, which is based on a dual-layered model, is presented. The military named entity group is divided into two categories: simple name and complicated ones. First the simple named entity is recognized in the first layer with CRFs, the results are transferred to the second layer to support the recognition of complicated named entity. In the second layer, SVM was used to decide the right boundary of a complicated named entity and then CRFs model was used to tag the former boundary. Last the simple and complicated named entity recognition results are combined together. Experimental results show that the model is effective.