This paper presents a novel technique for the enhancement of the operational reliability of processor arrays by a multi‐level fault‐tolerant design approach. The proposed fault tolerant architecture uses a flexible reconfiguration of redundant nodes, thereby offering a better spare utilization than existing two‐level redundancy schemes. The spare nodes at each level can replace any of the failed primary nodes, not only at the same level but also those at the lower levels. The architecture can be adopted to increase the system reliability in Multi Chip Modules (MCMs). The main contributions of our work are the higher degree of fault tolerance, higher overall reliability, higher flexibility, and a better spare utilization.