We'll start with a short recap of the lattice Anderson model, with a focus on Minami's estimate and its applications. In particular it implies that, with high probability, the eigenvalues of the Anderson model are well-spaced. In the bulk of the talk i then describe a new approach towards such a level-spacing estimate which is more flexible than known methods.In particular it works for the continuum Anderson model. If the single-site probability distributions are sufficiently regular, then a Minami-type estimate can be obtained from such a level-spacing estimate.The talk is based on joint work with Alexander Elgart.