Abstract: To solve the unconstrained multilevel lot-sizing (UMLLS) problems, the harmonious particle swarm optimization (HPSO) algorithm, hybrid scatter search(HSS) algorithm and genetic algorithm integrated with repulsion operator (RGA) have been proposed. To study their suitability, a performance comparison among them was carried out systematically via a comprehensive test introducing the standard testing data sets, then the alternative of those algorithms can be decided for UMLLS problems with different sizes. The test results showed that HSS is more powerful for small/medium-sized problems and HPSO is superior to the other two algorithms in large-sized problems.