The present study aims at comparing total cloud cover (TCC) as simulated by regional climate models (RCM) from CORDEX project with the same variable as simulated by the driving global climate models (GCM), which are part of the fifth phase of the Climate Model Intercomparison Project ensemble. The comparison is performed for the Mediterranean region, and for the 1971-2005 period, when results from the "historical" scenario can also be compared with two data sets of ground-based cloud observations. We work with 14 modeling results (resolution, 0.11 degrees x 0.11 degrees), which are a combination of five GCMs and five RCMs. In general, RCMs improve only very slightly the climatic estimation of TCC when compared with observations. Indeed, not all RCMs behave the same, and some indicators (monthly evolution of the relative bias) show an enhancement, while other indices (overall mean bias and annual range difference) improve only very slightly with respect to GCMs. Changes in the estimate of TCC in summer might be the most relevant value added by RCMs, as these should describe in a more proper way several mesoscale processes, which play a more relevant role in summer. Noticeably, RCMs are unable to capture the observed decadal trend in TCC. Thus, TCC simulated by RCMs is almost stable, in contradiction with observations and GCMs, which both show statistically significant decreasing trends in the Mediterranean area. This result is somewhat unsatisfactory, as if RCMs cannot reproduce past trends in TCC, their skill in projecting TCC into the future may be questioned.