The ability of the Santa Barbara Disort Atmospheric Radiative Transfer (SBDART) model to predict clear-sky diffuse and direct normal broadband shortwave irradiances is investigated. Model calculations of these quantities are compared with data from the Atmospheric Radiation Measurement program's Southern Great Plains (SGP) and North Slope of Alaska (NSA) sites. The model tends to consistently underestimate the direct normal irradiances at both sites by about 1%. In regard to clear-sky diffuse irradiance, the model overestimates this quantity at the SGP site in a manner similar to what has been observed in other studies <Halthore and Schwartz, 2000>. The difference between the diffuse SBDART calculations and Halthore and Schwartz's MODTRAN calculations is very small, thus demonstrating that SBDART performs similarly to MODTRAN. SBDART is then applied to the NSA site, and here it is found that the discrepancy between the model calculations and corrected diffuse measurements (corrected for daytime offsets) <Dutton et al., 2001> is 0.4 W/m2 when averaged over the 12 cases considered here. Two cases of diffuse measurements from a shaded black and white pyranometer are also compared with the calculations and the discrepancy is again minimal. Thus, it appears as if the diffuse discrepancy that exists at the SGP site does not exist at the NSA sites. We cannot yet explain why the model predicts diffuse radiation well at one site but not at the other.