This article presents an evaluation of four models that are widely used in developing Gaussian plume models, based on the atmospheric stability class and the Pasquill-Gifford curves. The four statistics used to evaluate the models are the sum of residuals, the minimized sum of the squared residuals, the correlation coefficient, and the estimated error standard deviation. Evaluation shows that all of the four models have problems in fitting the Pasquill-Gifford data when the downwind distance increases from the source. A new model has been developed which fits the Pasquill-Gifford data more effectively than the previously developed models studied.