Radiative heating is one of the principal drivers of tropical circulation. While we have good knowledge of radiative fluxes at the top-of-atmosphere and at specific surface sites, observations of atmospheric profiles of radiative heating, particular in cloudy conditions, have been largely unavailable. The Atmospheric Radiation Measurement (ARM) Program has begun a program to compute radiative heating profiles routinely at its observational sites at Nauru and Manus Island, Papua New Guinea, using observed and retrieved inputs of water vapor and condensed water phase, particle size, and mass. The accuracy of these profiles can be assessed by comparing the calculated TOA and surface fluxes with observations. We have computed radiative heating profiles every 20 minutes for several months at each of these two sites in the 1999-2000 time period, which represent a unique dataset for model comparison. Here, we compare this dataset to model output from the European Center for Medium-Range Weather Forecasting (ECMWF) analysis, the NCAR Community Atmosphere Model (CAM 3.0) and the Multi-Scale Modeling Framework (MMF). These three models, all of which are run using observed SST for this comparison, provide an interesting range of resolution from the 4 km cloud resolving model in the MMF to the approximately 280 km grid-scale of the CAM and a contrast between forecasting and climate models. In general, the model results provide a reasonable representation of the observed heating, but do not capture the structure of the observed heating in the upper troposphere because of their failure to simulate cirrus and stratiform cloud adequately.