The diurnal temperature range (DTR), defined as the difference between the maximum and minimum temperature, is a useful diagnostic index for evaluating global climate models. In this study, the DTR from 17 GCMs available in the fifth phase of the Coupled Model Intercomparison Project (CMIP5) is evaluated on the Tibetan Plateau (TP) by comparison with the observations during 1961–2005. During the studied period, the observed maximum/minimum temperatures on the TP show statistically increasing trends with the annual rates of 0.19/0.36 °C decade−1, respectively, leading to the reduction of DTR (−0.22 °C decade−1). Compared with the observed DTR, most CMIP5 models generally underestimate DTR, with absolute error ranging from −4.58 °C (GFDL‐ESM2M) to −1.36 °C (CESM1‐BGC). Fifteen CMIP5 models have reproduced the overall negative trends of DTR on the TP, but their trend magnitudes are smaller. Furthermore, the CMIP5 model biases in DTR are investigated by means of correlative approach, to reveal the dominant variables. The differences between the surface downwelling short‐wave radiation (SDSR) with clear skies (SDSRcs) and the SDSR can be used to describe the surface short‐wave cloud radiative effect. Similarly, the differences between the surface downwelling long‐wave radiation (SDLR) and the SDLR with clear skies (SDLRcs) are defined to address the surface long‐wave cloud radiative effect. It is found that the mean DTR in the CMIP5 models has significantly negative correlations with both SDLR–SDLRcs and SDSRcs–SDSR, suggesting that the model differences in DTR on the TP are probably determined by radiation variables and total cloud fraction in the CMIP5 models.