Abstract: The energy budget of two versions of GOALS model (GOALS-1.1 and GOALS-2) is described, and compared to the observational estimates. The results illustrate that the simulated net surface shortwave radiation flux is underestimated in the high latitude regions while the net longwave radiation flux is substant ially overestimated in that region, which results in the lower surface air temperature (SAT) in the polar region and the strongly negative sensible heat flux in the high latitudes. The overest imated sensible heat flux from surface to atmosphere in continents causes the much warmer SAT centers, which are the reasons for the bias of model SAT. The reasons that the simulated precipitation in most of the regions is less than observation are closely related to the underestimated latent heat flux over most of Eurasian continent and the oceans, especially over the subtropical ocean. It can be seen that the bias in the OLR of the two models lies in mid-latitudes and low latitudes, where the simulated solar absorbed shortwave radiation flux at top of atmosphere is comparable to NCEP reanalysis, but much less than ERBE data. This indicates that the improvement of cloud-radiation parameterized scheme in mid-latitudes and low lat itudes is of critical importance to the simulation of global energy budget. The simulated cloud cover from GOALS-2 model with diagnosed cloud scheme is generally less except equator areas, especially in the mid-latitude areas, which causes the large bias of energy budget there. This suggests that the refinement of cloud parameterization is one of the most important tasks in the model's future development.