FOI has developed a new dispersion model, "PUMA", based on a Gaussian puff approach. PUMA is a part of a program concept called Dispersion Engine, containing different dispersion models. The basic concept is that Dispersion Engine has a welldefined interface that allows it to be connected to other computer systems that can take advantage of the dispersion calculations. Part of the development of the model and the program concept has been to implement a validation and evaluation work which is presented in this report. Methods for evaluation of dispersion models are well documented (see references below), and is followed in this study. The experimental results have been compared with the calculation results from PUMA and evaluated using statistical methods. Such an experiment, which results from PUMA have been compared to, is known as "Prairie Grass" and allows for comparison of different levels of atmospheric stability. A second experiment is carried out by INERIS in France and was developed to study heavy gas behaviour. PUMA has a newly developed heavy gas component, which has been evaluated against the INERIS data. MSB provides a dispersion model through their decision support package RIB, called Spridning Luft. Spridning Luft version 1.4.2 is also included in part in this study in a comparison between models. Some parameters are difficult to identify by measurements in the atmosphere and can therefore provide an uncertainty in how these should be assessed. This is also apparent when you study the information about the experiment Prairie Grass. The uncertainty and even the effect of the variation of a parameter is large for the determination of the atmospheric stability and can sometimes be impossible to assess. In the report this issue is given extra attention. In a direct comparison between the measurements of the dispersion in the atmosphere and a good dissemination model, it can statistically be expected that 50% or more of the calculated values are within a factor of 2 from the measured value. It is also what we see in most of the comparisons that are carried out in this report. Thus, it is also possible to assess the calculation results from PUMA as acceptable. The result is most favourable for neutral stability while in the stable and unstable cases the spread in the measurements is larger and thus also the results for the comparison. In this study, a comparison between calculation results from PUMA and experimental data from two different experiments carried out. The variation in the results at individual comparisons are large, as also the variation in the wind field is large. Comparisons with further experiments is therefore important to clarify whether dissemination of the results of the PUMA statistically valid also in other cases.