The Nested Air Quality Prediction Modeling System (NAQPMS) has been applied to the routine air quality forecast in Beijing during the Olympic Games. Monte Carlo method is used to analyze the uncertainty of ozone simulation of NAQPMS during the Olympic Games, from 8 to 24 Aug 2008. Latin hypercube sampling has been used for multi-variables sampling, and 50 ensemble runs have been made with 154 parameter uncertainties being considered together. By the temporal average, the most important parameter to the surface ozone output uncertainty in Beijing is the local precursor emissions during the day time. Other important factors include NO2 photolysis coefficient, wind direction, precursor emissions from the surrounding areas of Beijing, and vertical diffusion coefficient. The wind direction and precursor emissions from the surrounding areas of Beijing have the greatest impact on the uncertainty of daytime ozone simulation at higher levels (above about 150 m). The main uncertainty factors in ozone simulation at night are local NOx emissions and vertical diffusion coefficient. Given the predefined input uncertainties, the average uncertainty of ozone simulation is 19 ppb, ranging from 2 ppb to 49 ppb.