One of the problems or challenges in performing graph clustering is to determine the number of clusters that best fit to the data being processed. This study is proposing a method to solve the problem using Dirichlet Process Mixture Model (DPMM). DPMM is one of the statistical methods that is already used for data clustering, without the need to define the number of clusters. However, this method has never been used before for graph clustering. Therefore, this study proposes the adaptation so that DPMM can be used for graph clustering. Experiment result shows DPMM method can be used for graph clustering, by applying spectral theory.