Recent advances in computer power have significantly decreased computational time and increased memory capacity. Consequently real-time numerical weather prediction has spread rapidly from operational centers such as NCEP to universities, government agencies, and private industry. The Weather Research and Forecast (WRF) model has been applied to the real-time forecast over southwestern United States. The model is configured with the center located at El Paso, Texas, with the spatial resolution of 27 km and 9 km for mother and nested domain, and with the AVN grib data as the initial and lateral boundary conditions. A one month long experiment provides valuable information about the performance of the WRF. This paper will present some statistical results based on the analyses of model forecast variables of surface temperature, wind speed as well as precipitation score (Threat Score and Bias). The successful implementation of the WRF model for southwestern U.S cities will provide more accurate local weather forecast and enhanced data input for air quality models in the Southwest.