Abstract: The method of characteristics (MOC) is usually solved with flat source approximation, which requires a large number of smaller meshes to obtain accurate results, which will consume more memory and affect the calculation efficiency. Compared with the flat source approximation, the linear source approximation can get more precise numerical results with fewer meshes. In this paper, a linear source MOC based on least square was proposed and analyzed by calculating pressurized water reactor (PWR) cell and C5G7 benchmark. The numerical results indicate that the linear source MOC based on least square can contribute stable and accurate numerical results with fewer meshes.