Dynamic light scattering is a powerful tool for measuring the size of nanoscale particles. However, the average particle size inverted by the traditional linear cumulants method depends closely on the length of correlation data. Linear and non-linear fitting algorithms are analyzed in order to overcome this disadvantage. An optimized fitting algorithm for the cumulants method is proposed based on the advantages of both fitting algorithms. The particle diameter is obtained from a first-order curve fit, and the particle distribution′s polydispersity from a second-order polynomial fit over the optimal range of the intensity correlation function. Theoretical analysis and experimental data show that the relative error and repeatability of the inverted diameter are less than 2%, and the relative error of polydispersity index is less than 6%. In conclusion, the optimal fitting algorithm for the cumulants method can be used to measure a stable and reliable particle size and its polydispersity index.