Atomic friction involves objects whose dynamics is strongly influenced by thermal fluctuations. In stochastic modeling, one focuses on a few relevant degrees of freedom, whereas the atomistic ones are taken into account by introducing dissipation and noise. We review applications of this approach to atomic friction, namely, the basic Prandtl-Thomlinson model, some of its multidimensional generalizations, and the rate approximation, which allows one to obtain analytical results not easily accessible by other methods.