Malware are now developed at an industrial scale and human analysts need automatic tools to help them. We propose here to present the results of our experiments on this difficult problem: how to cluster a very large set of malware (with only static information) to be able to classify some