This study provides an overview of the approach used to develop and validate new CERES Angular Distribution Models (ADMs) for estimating top-of-atmosphere (TOA) radiative fluxes from measured radiances. The new ADMs are determined empirically using nine months of coincident CERES and Visible Infrared Scanner (VIRS) measurements from the Tropical Rainfall Measuring Mission (TRMM) spacecraft. The CERES data include measurements from cross-track, alongtrack and Rotating Azimuth Plane (RAP) scan modes. In RAPS mode, CERES rotates in azimuth as it scans in elevation, thereby optimizing angular sampling. ADMs are determined for scene types defined according to parameters that have a strong influence on the anisotropy of Earth scenes (e.g. cloud optical depth, cloud fraction etc.). Several hundred ADM scene types are defined for different combinations of imager-based retrievals and atmospheric state parameters provided by ECMWF data assimilation analysis. The advantages of combining CERES and imager data are discussed, and accuracies in TOA fluxes from the new CERES/TRMM ADMs are compared with those from the CERES ERBE-Like product. CERES ERBE-Like TOA fluxes are based on ADMs developed during ERBE and rely on broadband radiances for scene identification. RMS errors in regional TOA fluxes from the new CERES ADMs remain less than 0.6 W m^-2 (daily average), corresponding to a 60% reduction compared to RMS errors from the ERBE-Like product. When stratified by viewing geometry, mean fluxes from the new CERES ADMs are consistent to within 2% (or 2 W m^-2) in the SW, and 0.6% (or 1.5 W m^-2) in the LW. In contrast, ERBE-Like mean fluxes show a systematic dependence on viewing zenith angle of approximately 10% (or 10 W m^-2) in the SW, and 3.5% (or 9 W m^-2) in the LW. Multiangle consistency checks show that instantaneous flux errors from the new CERES ADMs are smaller than those from the ERBE ADMs by up to a factor of 4, depending on scene type.