For the design optimization of the Pelton turbine, it is highly demanded to investigate the flow inside the turbine casing, which includes the water jet from the nozzles, the interaction between the jet and rotating runner, and the flow ejected from the bucket outlet, in various operating ranges. The behavior of the flow, however, is very complicated due to its unsteadiness and free surface. Further, the experimental observation in the model test is challenging because of the high time resolution requirement and the obstruction by the splashing water inside the turbine housing. In this regard, the numerical analysis is considered as a powerful method to approach the flow behavior in the Pelton turbine. Conventionally, the grid-based numerical analysis is applied to the calculation of the flow for its practical design. However, with the gird-based methods, a huge amount of fine elements is required to capture the unsteady behavior of the water free surface and the tiny splashing water particles. Increasing the number of elements directly results in higher computational costs, which makes it difficult to consider the splashing water in the process of practical design. From this point of view, the GPU-accelerated finite volume particle method is applied to the investigation of the flow in this research. It is firstly confirmed that the particle-based numerical result has a good agreement with the experimental result by the comparison of the turbine characteristics. Furthermore, several evaluations of the flow based on the analysis for the practical design are introduced in the paper.