The relevance of two rebound modeling approaches classically used in rockfall simulation codes was assessed using field experiments of single rebounds. A lumped mass model, modeling the rock as a single material point, and a rigid body one, explicitly accounting for the rock shape, were used. Both of them are efficient with only a few calibration parameters. The main limitations of each approach are the calibration of a parameter accounting for both the roughness of the soil and the rock shape, for the lumped mass approach, and the estimation of the rock length and height, for the rigid body approach. Finally, both rebound models require being improved to better predict the rotational velocities distribution.