Building phylogenetic trees is one of the significant applications within bioinformatics, mainly due to its involvement in multiple sequence alignment. Because of the high computational complexity required, the use of parallel processing during the building process is convenient. Taking into account that current cluster architectures are hybrid, in this paper we present a parallel algorithm to build phylogenetic trees based on the Neighbor-Joining method, which uses a hybrid communication model (combination of message passing and shared memory), and then analyze its performance. Finally, conclusions and possible future lines of work are presentedFacultad de Informática