The paper analyzes different version of the mini-model method (MM-method) which are based on centroid clustering algorithms. The article introduces two new versions of the method which are based on k-medoids method and the fuzzy clustering methods -c-means algorithm. The MM-method is an instance-based learning algorithm. It operates only on data from the local neighborhood of a query. The local neighborhood is created by clustering algorithms. It makes the learning procedure simpler, because in the previous version, the neighborhood of query was based on multidimensional polytopes. The article also presents comparison between different versions of the MM-method and other instance-based learning algorithms.