The present paper proposes a hybrid version of a compact genetic algorithm (cGA) as approach to solve the Multi-Level Capacitated Lot Sizing Problem. A fix and optimize heuristic is embedded with cGA to improve solutions while mathematical programming technique is used to evaluate them. This hybrid approach is tested over two sets of benchmark instances. The results achieved are compared with two time-oriented decomposition heuristics from literature and with a hybrid multi-population genetic algorithm recently proposed for the same problem. Computational results show that the hybrid cGA has a superior performance mainly for instances dealing with setup times.