Using the Spectral Atmosphere Model (SAMIL), developed by the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences (LASG/IAP), this study examines the simulation of radiative fluxes, by comparing with the observation and atmospheric multi-model results from IPCC AR4 (Intergovernmental Panel on Climate Change Fourth Assessment Report). In the old version of SAMIL, large biases of radiative fluxes are noted. The simulated radiative fluxes have been evidently improved in the new version of SAMIL, due to tuning the physics processes, i.e. radiation, cumulus convection, and cloud physics processes. Annual variation and mean of global radiative fluxes at the top of atmosphere are closer to the observation than those in the old version. The cloud physics process is further tuned on the basis of the adjustment of cumulus convection parameterization and then the new SAMIL reasonably represents the cloud properties. Meanwhile, the mismatch between cloud and radiation induced by only tuning the cumulus convection process in the old SAMIL is largely corrected. The new SAMIL could reproduce the basic feature of global climatological radiative fluxes, especially in East Asia. Furthermore, the results show that the biases of cumulus convection process largely affect radiative fluxes in the tropical and subtropical regions with active convection, and the simple cloud diagnosis scheme is one source of the major simulation biases. In summary, the cloud physics process in the SAMIL should be improved to better simulate cloud properties and physics process, and further cooperative adjustment is necessary, according to the study of interaction between cumulus convection and cloud physics processes.