The present paper is the third and last part of an investigation on what determines reliability in fuel cell model identification. In continuation to the effect of experimental design (Part I) and a process method for stochastic calculation of a model's parameters (Part II), this paper concentrates on the assessment of a model validation. Four criteria are examined. The fit of the model's output to experimental data, the determinant of the covariance matrix of the parameters, the determinant of their correlation matrix, and the product of their variances. As regards the fit to the data, results show that this is mainly a function of the number of measurement points. Repetitions do not seem to improve the average of the fit significantly, but it does improve its variation. For the other three criteria, which are also mathematically linked, results show a counterbalance between them, leading to the conclusion that they cannot be optimized simultaneously. This happens especially between the determinants of the covariance and the correlation matrices.