Recent studies have demonstrated that the performance of a simulated annealing algorithm can be improved by following multiple-search paths and parallel computation. In this paper, we use these strategies to solve a comprehensive mathematical model for a flexible flowshop lot streaming problem. In the flexible flowshop environment, a number of jobs will be processed in several consecutive production stages, and each stage may involve a certain number of parallel machines that may not be identical. Each job has to be split into several unequal sublots by following the concept of lot streaming. The sublots are to be processed in the order of the stages, and sublots of certain products may skip some stages. This complex problem also incorporates sequence-dependent setup times, the anticipatory or nonanticipatory nature of setups, release dates for machines, and machine eligibility. Numerical examples are presented to demonstrate the effectiveness of lot streaming in hybrid flowshops, the performance of the proposed simulated annealing algorithm, and the improvements achieved using parallel computation.