Abstract: The ability of 28 CMIP6 (Coupled Model Intercomparison Project 6) models that simulate the interannual variation and change of the climate mean state of the Diurnal Temperature Range (DTR) in China and different regional and seasonal scales was evaluated by using the CRU_TS v4.04 observation data as the benchmark. The results showed that the CMIP6 models can reflect the declining trend of the DTR at a centennial time scale in the interannual variation. The correlation coefficient between the model and the observation is 0.1–0.7, root-mean-square error is 0.6–1.5, and Taylor Score (TS) is 0.2–0.7. The correlation coefficient between the MRI-ESM2-0 model and the observation is the highest (0.65), root-mean-square error (0.8) is the lowest, and TS (0.67) is the highest. This indicates that the MRI-ESM2-0 model has the best simulation ability. At a 30-year climate mean scale, the CMIP6 models accord with the observed spatial distribution characteristics of the DTR, which is high in northern China, low in southern China, high in western China, high in eastern China, high in inland China, low in coastal areas, high in the plateau, and low in the plain basin. CMIP6 models can basically reproduce the declining trend over a large area of China in the climate mean state, and the DTR variation in different regions and seasons are also well simulated, with the EC-Earth3 model exhibiting the best performance. However, the individual model is easy to overestimate or underestimate the DTR variation to some extent. The multi-model ensemble can simulate some characteristics of the DTR in the interannual variation and change of the climate mean state, which is better than the single model for the spring and winter simulation.