Group technology (GT) layout is a layout that combiines advantages from product layout and process layout. cell formation is the frist step to apply it. in this research, particle swarm optimization (PSO) algorithm is used for machine grouping. while degree of belongingness is used for grouping prat families. the considered constraints are machine capacity , propesing time, and demand of each part. performance measure used as the objective function is generalized grouping efficacy. the model is implemented in caces which differences in number of machines, number of components, domand of each component, and processing times. each case is tested with several combinations number of parrameter wmax (inertia weight maximum), cl (self confidence), c2(social confidance. the results shows that matrix size affect sensitivity model to parameter. smaller matrix is not sensitive to PSO parameters. larger matrix is sensitive to parameter c, where using large c results in difficulty to reach convergence but reach better global best faster than using small c. parameters c1 = 1 and c2 = 2 show less iterations needed to reach global best. based on the implementation, it can be shown that the model has better performance comprared with tabu search and genetic algorithm.