Hydrological models have to simulate the stream flow of a river basin based on available hydro-meteorological data and other significant parameters. Semi-distributed and physically based watershed models are a class of deterministic hydrological models. For a specific watershed, selecting the most suitable hydrological model is necessary to obtain good simulated results. The main objective of this study was to evaluate the performance of HEC-HMS and SWAT hydrological models for simulating stream flow in the Gumara River Basin, in Ethiopia. In this thesis work, 12 years (2003–2014) of daily hydro-meteorological data were used. Analysis of hydro-meteorological data, sensitivity analysis, calibration, and validation was carried out in this study to evaluate the two hydrological model performances. The performances of these two models were compared to select the most suitable hydrological model for the study basin. Both models were calibrated and validated with observed stream flow data for 7 years (2004–2010) and 4 years (2011–2014), respectively. Coefficient of determination (R2), Nash–Sutcliffe efficiency (NSE), and percent of bias (PBIAS) were used to evaluate the efficiency of the models. The results of calibration and validation indicated that, for the Gumara River Basin, both models could simulate the stream flow. In the SWAT model, the R2 = 0.80, NSE = 0.79, PBIAS = 3.2%, and R2 = 0.75, NSE = 0.70, and PBIAS = 23.7% for the calibration and validation period, respectively. The HEC-HMS model provided the model performance for the calibration period with the R2 = 0.55, NSE = 0.509, and PBIAS = − 12.91%, and for validation periods with the R2 = 0.54, NSE = 0.32, and PBIAS = 26% in daily time steps. The simulation obtained by the SWAT model is more satisfactory than the HEC-HMS model for stream flow in the Gumara watershed.