In combustion applications, the weighted-sum-of-grey-gases (WSGG) model is widely used because it is computationally efficient, but it often yields relatively large errors. The objective of this study is to examine the effect of radiation models on CFD predictions of flame spread. To this end, a statistical narrow band (SNB) model and the WSGG model are employed for the simulation of two upward flame spread scenarios, one being a largescale flame spread over a vertical PMMA wall while the other representing flame spread along vertical corner walls. The entire flame spread model also consists of a four equation turbulence model, an eddy-break-up (EBU) combustion model, a discrete transfer (DT) radiation model, a simple soot model and a non-charring pyrolysis model. Quantitative comparison is made between the prediction results obtained with the SNB model and the WSGG model as well as the experimental data. Results clearly show that the SNB model yields more accurate results than the WSGG approach. However, the SNB model is about four to five times more time consuming than the WSGG model. Therefore, for simulations of complex engineering applications a compromise between accuracy and numerical efficiency should be taken into account.