A comparison between two Artificial Intelligence (IA) optimization methods and a traditional gradient based method, applied to a model predictive controller (CPBM) is shown. The model used is non-linear and it includes the effect of a real pneumatic valve type Final Control Element (EFC), because that effect tends to adversely affect the performance of the CPBM when it's model has an ideal EFC. The CPBM application uses Genetic Algorithms and Bacterial Chemotaxis as IA techniques. The effectiveness of the techniques for a real implantation as well as the potential for IA techniques for CPBM are explored. The simulation uses an existent continuous stirred tank reactor (CSTR) benchmark problem.