Simulating clouds and their radiative response remains one of the greatest challenges in climate models due to the complexity of different cloud types. In this study, we evaluate the individual cloud feedbacks in CMIP6 models by comparing them with the CERES observations. Following Zelinka et al. (2022), we decompose clouds into different components by changes in cloud amount, altitude, and optical depth. The clouds are further separated into low cloud and non-low cloud over land, and ocean ascent descent regions. The observations are available from 2002 to present, so we focus on the feedbacks in response to interannual variability. These observed cloud feedbacks are then compared to feedbacks inferred from CMIP6 fully coupled and AMIP model runs. The results suggest tropical marine low cloud feedback and high cloud altitude feedback are the two major components that lead to the discrepancies between observations and models. The comparison provides insights into the cloud feedback components that need to be improved in models and motivates future study.