Redundancy Allocation Problem (RAP) is one of the most important problems in the eld of reliability. This problem is aimed at increasing system reliability under constraints such as cost, weight, etc. This study works on a system with series-parallel con guration and multi-state components. To draw the problem nearer to the real condition, this study merges this problem with discount levels in purchasing components. For calculating the reliability of sub-systems, a recursive algorithm is used. Because the redundancy allocation problem belongs to NP-hard problems, for optimizing the presented model, a new Genetic Algorithm (GA) was used. The algorithm parameters were tuned using Response Surface Methodology (RSM), and an enumeration method was used for the validation of GA.