In the present work, we have assessed the circumsolar version of the simplified Perez model which estimates the illuminance received on vertical surfaces;facing N,E,S and W, using mean hourly values of illuminance obtained at Madrid for a period of 13 consecutive months. The coefficients of the circumsolar version have been determined in two different ways: using jointly the data obtained for the four orientations, or using for each orientation the data obtained specifically for it. In the last case, the circumsolar version has been assessed for six different values of the circumsolar half angle alpha = 7.5 degrees, 15 degrees, 20 degrees, 25 degrees and 35 degrees. Although the best estimation (RMSD) is obtained generally for alpha = 25 degrees, similar deviations are obtained for all angles excepting alpha = 7.5 degrees. It has been verified that, with Madrid＇s data, the circumsolar version performs much better when coefficients are determined independently for each orientation than when using jointly the data obtained for the four orientations. The results obtained with the circumsolar version of the Perez model have been compared with those given in a previous work [3] (Robledo, L. and Soler, A., Lighting Res. Technol., 1996, 28(3), 141.), where the point source version of the Perez model was assessed using the same data sets. The prediction accuracy obtained with the cicumsolar version using alpha= 25 degrees and coefficients calculated specifically for each orientation is similar to that obtained in the mentioned work when considering different sets of coefficients for the Vertical surfaces seeing or not seeing the centre of the solar disk.