The all-sky Aerosol Direct Radiative Effect (ADRE) varies considerably among global climate models (GCMs), which results from differences in aerosol and atmospheric states and ADRE sensitivity to aerosol-related radiative processes. This study uses a regression method to analyze the inter-model spread of ADRE among the GCMs of the Sixth Coupled Model Intercomparison Projects (CMIP6). The key state variables examined include scattering and absorbing aerosol optical depth, surface albedo, and shortwave cloud radiative effect. We find that differences in state variables and radiative sensitivity explain 67% and 17%, respectively, of the global ADRE anomaly. The ADRE anomaly in different models is dominated by different factors, which sometimes leads to compensating effects. For the global mean ADRE anomaly, aerosol optical depth differences dominate in CNRM-ESM2-1 and GFDL-ESM4 models, while ADRE sensitivity variations to aerosol-only scattering effect dominate in HadGEM3-GC31-LL, MPI-ESM-1-2-HAM, and MRI-ESM2-0 models. Aerosols scatter and absorb solar radiation, impacting the Earth's climate. Global climate models differ in their quantification of this effect. The differences in quantification arise from different state variables, including the aerosol properties and atmospheric conditions. Additionally, the radiative sensitivity of aerosol effect to aerosol-related physical processes varies with models. In this study, we identify the primary causes of the differences in aerosol radiative effect among climate models and quantify their respective impacts by using a regression method. We find that besides the aerosol optical depth, whose influence on aerosol radiative effect is well recognized, the radiative sensitivity of aerosol and surface interactions significantly contributes to the aerosol radiative effect differences in the latest climate models. A regression model is used to dissect the inter-model spread of the all-sky Aerosol Direct Radiative Effect (ADRE) in Sixth Coupled Model Intercomparison Projects ModelsThe model explains 86% of ADRE inter-model spread, with variances in state-variable and radiative sensitivity contributing 67% and 17% eachDifferences in aerosol optical depth and ADRE sensitivity to aerosol-surface interactions drive the spatial variance in global ADRE spread