In this paper, we propose a novel intelligent method to improve the calibration quality of parametric exponential Lévy models that have recently emerged as alternative option pricing models. The method based on so-called multi-basin systems consists of three sequential phases to expedite the search for a good parameter set and to reduce the burden of selecting proper initial set of particles for particle swarm intelligence techniques. We conduct simulations on model-generated option prices and real data of option prices to verify the performance of the proposed method and show that the method can significantly improve the calibration quality in a systematic and automatic way.