This paper focuses on quasi-bipartite synchronization of heterogeneous memristive neural networks (MNNs) with cooperative–competitive interactions. Firstly, MNNs with nonidentical uncertain parameters are investigated. A pinning control is adopted to study robust quasi-bipartite synchronization of uncertain MNNs over the signed graph. A sufficient condition is derived for robust quasi-bipartite synchronization. Secondly, a more general heterogeneous MNNs is further discussed over the signed graph. Due to the heterogeneities of the MNNs, quasi-bipartite synchronization is studied and the corresponding sufficient condition is obtained. Additionally, the upper bounds of quasi-bipartite synchronization for two kinds of MNNs are given. Finally, two examples are given to illustrate theoretical results.