The pretreatment neuroimaging markers from the resting-state brain network that could predict the early response to antidepressants are still unclear. The aim of the present study was to identify the performance of reward network features for discriminating patients with a dampened response to antidepressants. A total of 81 major depressive disorder (MDD) patients (44 patients with treatment-responsive depression (RD) and 37 patients with non-responding depression (NRD)) and 43 healthy controls (HC) underwent resting-state functional magnetic resonance imaging scans and clinical estimates. Bilateral nucleus accumbens (NAcc)-based networks were constructed for further functional connectivity (FC) analysis. The FC of the right superior frontal gyrus (SFG) (area under curve (AUC) = 0.837) and left parahippocampus (AUC = 0.770) within the left NAcc reward network, as well as the FC of the left SFG (AUC = 0.827) within the right NAcc reward network, could distinguish NRD subjects from RD subjects relatively well. Taken together, when considering the distinctive connectional pattern of the bilateral reward circuits, the synthetical differentiating effect was achieved to an optimal performance for discriminating NRD patients (AUC = 0.869), with balanced sensitivity (0.838) and specificity (0.818). The distinct pretreatment characteristics of the reward network make specific contributions to the early response to antidepressants and establish a promising imaging predictor for the classification of early efficacy.