Abstract <p>The stability of determining the particle size distributions using model data on the intensity of small-angle scattering from mixtures of polydisperse spherical and cylindrical particles has been analyzed. It is shown that, when using a parametric distribution model, the probability of obtaining the correct solution using the nonlinear least-squares method decreases, as the range of starting parameter values expands, and the spread of solutions corresponding to local minima of the target function increases. Ambiguity of solutions occurs in most cases with overestimations of the average size of cylindrical particles and their polydispersity compared to the exact solution. To avoid falling of the minimized target function within local minima, a good starting approximation is required, which is within 50% of the relative detuning of the model parameters from their exact values.</p>