The performances of 43 CMIP5 (Coupled Model Intercomparison Project Phrase 5） models which participate in the Fifth IPCC Assessment (IPCC AR5) in simulating the land surface temperature （0 cm） over Northeast China during 1981-2005 are compared and analyzed using the outputs of CMIP5 models and the observation data from 173 meteorological stations. Results show that：most the climate model simulation results can reproduce the temporal and spatial changes of land surface temperature in the study area; the time correlation coefficient of the monthly observation is higher than 095; the cold deviation is found mostly throughout the studied period, and the simulation capability of interannual variation is limited. The spatial distribution shows that the CMIP5 models can simulate the spatial temperature distribution characteristics of high in the south and low temperature in the north over Northeast China, but the differences between the simulation results of different models are large and the models can simulate the warm and cold centers. The difference between the good and poor model groups in summer is the largest, and the poor model group cannot reproduce the distribution characteristics of the surface temperature. It is found that the FGOALS_s2 model is optimal. In general, the CMIP5 coupled climate models have good performances for the temporal and spatial variation of surface temperature in Northeast China, and the simulation performance of monthly value is better than that of interannual variation.