The vertical distributions of cloud water content (CWC) and cloud fraction (CF) over the tropical oceans, produced by 13 coupled atmosphere-ocean models submitted to the Phase 5 of Coupled Model Intercomparison Project (CMIP5), are evaluated against CloudSat/CALIPSO observations as a function of large-scale parameters. Available CALIPSO simulator CF outputs are also examined. A diagnostic framework is developed to decompose the cloud simulation errors into large-scale errors, cloud parameterization errors and covariation errors. We find that the cloud parameterization errors contribute predominantly to the total errors for all models. The errors associated with large-scale temperature and moisture structures are relatively greater than those associated with large-scale midtropospheric vertical velocity and lower-level divergence. All models capture the separation of deep and shallow clouds in distinct large-scale regimes; however, the vertical structures of high/low clouds and their variations with large-scale parameters differ significantly from the observations. The CWCs associated with deep convective clouds simulated in most models do not reach as high in altitude as observed, and their magnitudes are generally weaker than CloudSat total CWC, which includes the contribution of precipitating condensates, but are close to CloudSat nonprecipitating CWC. All models reproduce maximum CF associated with convective detrainment, but CALIPSO simulator CFs generally agree better with CloudSat/CALIPSO combined retrieval than the model CFs, especially in the midtroposphere. Model simulated low clouds tend to have little variation with large-scale parameters except lower-troposphere stability, while the observed low cloud CWC, CF, and cloud top height vary consistently in all large-scale regimes.