The pressure drop is an essential performance index to evaluate and design gas cyclone. In order to accurately predict the complex non-linear relationships between pressure drop and geometrical dimensions, a multi-fidelity surrogate, Co-Kriging is developed and employed to model the pressure drop for cyclone separators using CFD simulations dataset and Muschelknautz model. Only the inlet section dimensions, the vortex finder diameter and the cone height are significant. The Co-Kriging model is more accurate than the ordinary Kriging model. This accuracy can be further enhanced if the optimal Latin hypercube replaced the random sampling plan. In this study, five approaches are proposed and tested for sampling update for the Co-Kriging which was not covered in the literature. These approaches succeeded to enhance the model accuracy by adding few points from both the low-fidelity and high-fidelity simulations. The Co-Kriging model has been used to get a new optimized cyclone for minimum pressure drop using the genetic algorithms optimization technique. A CFD comparison between the new design and the standard Stairmand design has been performed. CFD results show that the new cyclone design is very close to the Stairmand high efficiency design in the geometrical parameter ratio (except the cone height), and superior for low pressure drop. The new cyclone design results in 22% of the pressure drop obtained by the Stairmand design at the same volume flow rate. (C) 2014 Elsevier B.V. All rights reserved.