We present a simple workflow for 2D smooth modeling of Radio-Magnetotelluric (RMT) data using Particle Swarm Optimization (PSO). The workflow aims for the general problems of the Global Optimization (GO) methods, encountered during multidimensional geophysical modeling; time consumption, and difficulties in solving large amount of model parameters. In our implementation of PSO, starting models are used and an optimum parameter change vector (Delta m) is searched by the algorithm. Our initial studies showed, implementing a smooth modeling approach is necessary for modeling large amount of model parameters using PSO. To obtain smooth models and avoiding multi-objective optimization we used an operator to smooth the model parameters to the desired level. As result, we used the PSO to search for the Am minimizing the functional when the model parameters are smoothed. The smoothness is gradually decreased when rougher models are wanted. As the smoothness decrease, artefacts are found to be emerging; especially at the low sensitivity zones. To prevent these artefacts, we added Minimum Gradient Support (MGS) to our functional with a multiplier. Since the value of this multiplier is necessarily small, the algorithm doesn't search for a balance between these objectives, and the contribution from the MGS acts as a threshold value to introduce a structure to the model. We tested this approach using a synthetic and a field dataset In all cases the algorithm successfully minimized the functional and obtained geologically meaningful models. We also compared the uniqueness of the models recovered from the field dataset by multiple modeling realizations using smooth inversion and PSO. Our results suggest, when higher smoothness and minimum structures are aimed, the uniqueness of the models is significantly higher, however, for a fixed set of modeling parameters with moderate values, the non-uniqueness is observed to be higher than that suggested by the smooth inversions. (C) 2019 Elsevier B.V. All rights reserved.