Studying the differential characteristics and driving factors of land use efficiency in different functional cities under the condition of undesirable outputs has great significance for urban sustainable development. This paper applied the improved EBM model in DEA to measure land use efficiency for 268 Chinese cities across five categories. Grey correlation modeling was used to explore the main driving factors of land use efficiency. We found that ignoring ecologically negative effects leads to an overestimation of land use efficiency. Land use efficiency of each functional city was different, proving that the measurement was scientific in terms of city function type. Land use efficiency curves were relatively steep and had periodic variation without considering the undesirable output while this curve became relatively gentle and had less periodic variation when taking the undesirable output into consideration. The main driving factors of land use efficiency in different functional cities contain significant differences. For integrated cities these driving factors included industrial structure, degree of land marketization and capacity of environmental governance; for resource-based cities they were industrial structure, external dependence, government regulation, industrial structure; for industry-based cities they were the degree of land marketization and regional energy structure; for entertainment-based cities they were the level of urbanization, economic development and capacity of environmental governance; for local cities it was mainly affected by the level of urbanization, economic development and government regulation. We need to fully consider the undesirable output and city heterogeneity in order to improve urban land use efficiency and determine different land management policies.