Reliability-based design methods have been advocated in recent geotechnical design codes to rationalize the design process of deep foundations. A key factor in reliability analyses is the realistic evaluation and incorporation of model uncertainty in the methods used for calculating skin friction and end bearing. In this paper, a database of published axial load tests conducted on drilled shafts in sand, gravelly sands and gravels is used to quantify the model uncertainty for two well-known methods for predicting skin friction of drilled shafts in granular soils. Statistics of the model factors (ratio of measured to predicted skin friction) were determined and classified as a function of soil type and soil layer depth below the ground surface. Results of the model uncertainty analysis indicated that the available methods underpredict skin friction particularly at shallow depths. In addition, the model factors exhibited coefficients of variation ranging between 0.4 and 0.6, indicating a significant level of model uncertainty in the skin friction predictions. The model factors were then used to evaluate the reliability index that is associated with drilled shafts that are designed according to factors of safety ranging between 2 and 3. Results of the reliability analysis as obtained using Monte Carlo simulations reflected reliability indices that were between 2 and 4 for drilled shafts in gravelly sand and gravel. For cases involving drilled shafts in sand, reliability indices were found to range from 0.9 (long shafts designed at a factor of safety of 2.0) to reliability indices of 4.0 (short shafts at a factor of safety of 3.0). These results indicate that the factors of safety or resistance factors that are needed in design codes to yield constant target reliability indices for drilled shafts in granular soils are expected to be dependent on the type of granular soils and the design length of the shaft.