The definition of sensitivity analysis on Computational Fluid Dynamics (CFD) modeling for gas dispersion has been presented. It is an uncertainty analysis method during the modeling process which firstly identifies the most sensitive ones from a number of uncertain factors, and then monitors and analyzes their impacts on the simulation results, and lastly selects the most suitable modeling parameters. The advised orders for the analysis should be grid dependency, boundary conditions, turbulence models, and parameter analysis for solution controls, successively. The analysis theories and methods for two key factors including grid dependency and turbulence model sensitivity have been discussed. A case study of gas dispersion in a station for high sulfide hydrogen natural gas gathering has also been carried out. The results show that grid dependency analysis could obtain a balance point between model prediction accuracy and computation cost; most suitable turbulence description method could be selected by means of comparison between the prediction results made by both turbulence models and empirical formulas. Sensitivity analysis is important for establishing proper computation models, enhancing prediction accuracy, and reducing computation costs, and it is an indispensable step for gas dispersion CFD modeling.