Investigating the robustness of memristor-based neural networks (MNNs) under disturbances is crucial since robustness improves the practical effectiveness of neural networks. This article investigates the robustness and the multiple $\psi$-type stability for MNNs with unbounded time-varying delay under input perturbation. New criteria for boundedness and $\psi$-type stability of MNNs under input perturbation which satisfies some conditions are provided through pure analysis methods.