I outline the development of four generations of kinetic models, starting with Chamberlain's solar breeze exospheric model. It is shown why this first kinetic model did not give apposite supersonic evaporation velocities, like early hydrodynamic models of the solar wind. When a self-consistent polarization electric potential distribution is used in the coronal plasma, instead of the Pannekoek-Rosseland's one, supersonic bulk velocities are readily obtained in the second generation of kinetic models. It is outlined how the third and fourth generations of these models have improved the agreement with observations of slow and fast speed solar wind streams.