For statisticians, multi-level models are often called mixed models. The difference is extra latent random variables being conceptualized into the 'usual' model (e.g., logistic regression).
There are many challenges for the so-called multi-level models for non-gaussian responses. The integration over random effects are usually difficult. There are a few ways to do approximate inference, but don't be cheated by any software and believe that everything will work out nicely.
