This article uses gap metric method to design a multi-model controller for nonlinear systems. In order to decompose the nonlinear system into a reduced nominal local models bank as much as possible, and assure the closed-loop robust stability and performance, the decomposition and designing of local controllers are integrated. To this end, robust stability, performance, and gap metric are incorporated to build a binary distance matrix that supports defining the driving and dependence powers for each local model. Then a best–worst analysis is employed considering the driving and dependence powers to find out the nominal local models. The proposed approach screens the value of all local models to choose each nominal local model. As a result, the global multi-model controller has a simple structure and avoids the computational complexity issues. To evaluate the effectiveness of the proposed method, two highly nonlinear systems, pH neutralization and continuous stirred tank reactor process, are simulated.