The numerical stochastic Lagrangian modeling of the passive tracer in a convective atmospheric boundary layer (CABL) was performed based on the random walk and Langevinmodels of turbulent dispersion. The statistical structure of turbulence is modeled by the probability density function (PDF) of vertical velocity fluctuations, which is recovered by the calculated statistical moments of the vertical velocity fluctuations. Four models of the PDF reconstruction were tested and the results of simulations are compared with the experimental data in CABL. The superiority of Langevin model over the random-walk models is demonstrated.