For a hybrid flow shop (HFS), due to the maintenance, partial failure, and possibility of failure, the number of machines at a workstation presents multiple levels, meaning that capacity of each workstation is stochastic. This study considers the stochastic capacity of each workstation and proposes a performance index based on system reliability to measure the probability that the HFS can complete demand within time constraint . The HFS is modeled as a stochastic-flow network, in which each arc is regarded as a workstation with stochastic capacity and each node as a buffer. An algorithm is then developed to find the lower capacity vectors that satisfy and system reliability is evaluated. In addition, a practical example of an IC card manufacturing system is utilized to illustrate the proposed algorithm and show that such an index can provide enough information for managers to make decisions.