Regression trees are powerful Machine Learning models capable of both flexibility in modeling as well as interpretability when the tree is not too deep. The M5 algorithm was introduced by Quinlan in 1992 under the name "model tree" ; the algorithm is derived from classic regression trees (e.g. CART, Breiman et al., 1984), adding the possibility to prune the tree and use linear regression models at leaves. The goal is to reduce the number of branches and leaves in the tree, making it ultimately more interpretable and smooth. The M5 algorithm was improved by Wang & Witten in 1997, under the name M5 Prime (acronym M5' or M5P). The algorithm gained popularity in particular a dozen years later with the Weka Machine Learning toolbox, providing a java-based implementation. `python-m5p` is an implementation of the M5P algorithm compliant with scikit-learn.