Soil erosion, land degradation and loss of agricultural soils are a major problem in Upper Blue Nile Basin of Megech and Gilgel Abbay watershed, Ethiopia. The parameter efficient semi-distributed watershed model, soil and water assessment tool and Generalized watershed loading function are tested for the prediction capability of sediment yield in Megech and Upper Gilgel Abbay watersheds. The objective of this study was to evaluate sediment prediction capability of three hydrological models (PED-WM, SWAT and GWLF) in a range of watershed sizes in the Upper Blue Nile Basin. The selected watersheds were Megech (508Km 2 ) and Upper Gilgel Abbay (1660 Km 2 ). Model calibration and validation were carried out using suspended sediment yield for Upper Gilgel Abbay watersheds from 1997-2007 and 2008-2012, respectively and for Megech watershed from 2000-2010 and 2011-2014, respectively. For evaluation of the model, the method was used model performance evaluation criteria (NSE, R 2 and PBIAS) were used. The PED-W model efficiency on monthly time step for sediment yield simulation the value of R 2 , NSE and PBIAS was (0.89, 0.77, -8.5) and (0.82, 0.81, 5.87) were obtained for Upper Gilgel Abbay watershed and (0.85, 0.71, 6.54) and (0.83, 0.72, 12.1) were obtained for Megech watershed during calibration and validation period respectively. The SWAT model efficiency on monthly time step for sediment yield simulation the value of R 2 , NSE and PBIAS was (0.84, 0.84, -2.7) and (0.62, 0.62, 3.5) were obtained for Upper Gilgel Abbay watershed and (0.64, 0.63, 17.9) and (0.63, 0.60, 27.9) were obtained for Megech watershed during calibration and validation period respectively. The GWLF model efficiency on monthly time step for sediment yield simulation the value of R 2 , NSE and PBIAS was (0.76, 0.58, -21.24) and (0.76, 0.60, -5.2) were obtained for Upper Gilgel Abbay watershed and (0.76, 0.57, 22.42) and (0.73, 0.58, 20.89) were obtained for Megech watershed during calibration and validation period respectively. Except PBIAS value for Megech watershed during validation of SWAT model all other model performance measure such as NSE, R 2 and PBIAS result for the two watersheds were the range of satisfactory to a very good agreement between the observed and simulated suspended sediment yield in the selected models. The overall model performance indicated that PED-WM model was more appropriate model to predict sediment yield than SWAT and GWLF model. Key words: Hydrological model, Gilgel Abbay, PED-WM, SWAT, GWLF, Megech.