Arid and semi-arid regions of the world are confronted with limited water resources. A large part of Iran is arid and semi-arid and rainfall in arid and semi-arid regions is typically meager, irregular and highly variable. This irregularity affects the hydrological cycle and water resources. Investigating the hydrology of the arid and semi arid regions is essential to know this environment and determine their vulnerability to changes. This is obvious that effective water resource management is necessary and this needs a decision support system that includes modeling tools. Choosing a model needs recognition of capability and limitations of hydrological models in watershed scale. In this paper for runoff simulation in semi-arid Azam Harat river basin, three conceptual continuous Rainfall–Runoff models HBV, HEC-HMS and IHACRES were used. HBV (Hydrologiska Byrans Vattenavdelning) model was firstly developed in Swedish meteorlogical and hydrological center in 1976. Up to now, the runoff simulations of different basins with different hydrological conditions have been evaluated by this model. This model simulates the continous runoff as well as flood single event of a basin, dividing the basin into several subbasins. Dividing subbasins is based on altitude and the vegetation of the basin. In this research we used the HBV-Light version. In this version Genetic Algorithm (GA) procedure is used to calibrate the parameters of the model. HEC-HMS (Hydrologic Engineering Center- Hydrologic Modeling System) model is a new version of HEC-1 model which has been used for simulation of both continous and single event runoff of a basin. On of the main advantage of this model is simulating the snow melt of the basin. In this research, the soil moisture algorithm was chosen, as the main methodoly of simulating runoff base on the fluctuations of rainfall, evapotranspiration and soil moisture losses. IHACRES model is based on non-linear loss module and linear unit hydrograph module. The process of simulation includes converting precipitation and temperature in each time step to effective rainfall by non-linear module, then converting to surface runoff by unit hydrographs linear modulus at the same time step.Some criteria of evaluation in this study are Nash coefficient (E), coefficient of determination (R2), and the standard error of a root mean square error (RMSE) and Bias. The results show that HBV model with 0.76 Nash coefficient, 0.77 coefficient of determination, 0.72 RMSE and -0.004 Bias error and HEC-HMS with 0.62 Nash coefficient, 0.64 coefficient of determination and 1.3 RMSE and 0.007 Bias error have highest and lowest efficiencies in the calibration period, respectively. These values are 0.66, 0.67, 0.8 and -0.15 for HBV model and 0.55, 0.57, 1.02 and -0.03 for HEC-HMS model, respectively. Finally HBV model has the best performance in simulating rainfall according to watershed condition in the validation period. In parameter sensitivity analysis that was applied, the most sensitive parameters of HBV model were UZL, mAXBAS and BETA. In HEC-HMS model, parameters soil storage, Max infiltration and tension storage were the most sensitive parameters with greatest effect on the model output results. The parameters of IHACRES model demonstrate equal sensitivity.