OptimOTU calculates single-linkage clusters at multiple thresholds given a distance matrix, which may be sparse, using a fast, multithreaded C++ implementation. In particular, the memory requirements are small and fixed, and it can accept the distance matrix from a file or text connection, meaning that it can operate on very large matrices which do not fit in RAM. Routines for quickly counting the measure of the intersection of sorted sets, and for calculating the multiclass F-measure, are also included.