Newer dynamic load models are being developed in industry through field-tests, to better represent the actual system response. Some of these models are quite complex, and include feeders, distribution transformers and various types of static and dynamic loads, thus resulting in more than a hundred parameters per model. This paper presents a methodology on how such complex dynamic load models in commercial software packages can be tested for errors such as software bugs, or implementation differences across the packages. The key lies in testing individual components of such models, through comparisons of simulation results from different packages. This method could be particularly useful where actual PMU data is not available for validating parts of load models. Case studies show that implementation differences exist due to incomplete documentation or bugs, and the techniques in this paper can be useful in finding and addressing the discrepancies in these evolving load models.