Clouds in the atmosphere reflect incoming shortwave radiation, trap outgoing longwave radiation, modulate precipitation, and consequently regulate the surface temperature. However, confidence in cloud simulations in global climate models (GCMs) remains low, and evaluation of their characteristics aiming for model improvement is actively pursued. Here, we investigate cloud diurnal variation (CDV), an aspect that so far has not been adequately addressed in Coupled Model Inter-comparison Projects (CMIP). The study was conducted by using a diagnostic parameter, the “effective-daytime cloud fraction” which accounts for the concurrent variation of clouds and solar insolation. When compared with ISCCP data, the majority of 20 CMIP5 models show good agreement over oceans, but significant biases exist over land (notably deserts and plateaus) attributable to smaller daytime cloud fractions. The CDV bias, while consistent with known bias in model simulated shortwave cloud radiative effect, also explains the inter-model differences.