Based on the radiation evidence from the monthly ISCCP (International Satellite Cloud Climatology Project) data, the performances of three state-of-the-art Atmospheric General Circulation Models (AGCMs), namely BCC AGCM, IAP SAMIL, and IAP GAMIL, in simulating the cloud radiative forcings (CRFs) are evaluated in terms of climatology and the response to ENSO. It shows that all of the three models can reasonably simulate the spatial patterns of annual mean long-wave and short-wave cloud radiative forcings and net cloud radiative forcing. The spatial pattern correlation coefficients between the model results and ISCCP data are all statistically significant at the 1% level. However, the differences between models results and ISCCP data in the magnitude of cloud radiative forcings are significant over most of the regions, especially over the subtropical oceans. In addition, there exist significant discrepancies in cloud radiative forcings over the off-equatorial central northern Pacific as well as over the southeastern and southwestern Pacific in the three models. Further analysis shows that the discrepancies in cloud radiative forcings are caused by unrealistic cloud fraction and the vertical structure of cloud which are determined by the convection bias and choice of cloud parameterizations. The spatial patterns of the responses of net cloud radiative forcing to ENSO are well reproduced by the three models, except that the magnitude of the negative response over the tropical western Pacific is overestimated (underestimated) by BCC AGCM and IAP SAMIL (IAP GAMIL), and the location of negative response in IAP GAMIL shifts westward to the tropical central Pacific.