Genetic algorithms (GAs) have shown to be suitable for optimizing real value functions with a considerably large number of variables. Thus, we have applied GA to develop an equivalent-circuit or empirical model optimization technique of microwave devices. The result of the optimization will provide accurate device modeling for geometric and material parameters chosen and therefore, models computationally very efficient with respect to any electromagnetic technique. This technique is illustrated for a tunable microwave phase shifter whose dispersion model has been optimized for a fixed structure.