There is evidence that obesity or higher body mass index is correlated with cognitive impairment in schizophrenia. Recent studies have demonstrated that genetic risk factors, such as the NRG3, are correlated with both elevated BMI and reduced cognitive function. In present study, we aimed to determine whether possession of the NRG3 rs10748842 influences the correlation between elevated BMI and reduced cognitive ability in schizophrenia. To our knowledge, this has never been examined before. A total of 625 inpatients with schizophrenia and 400 controls were recruited. The Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) was performed to assess cognitive function. We used multiple analysis of covariance (MANCOVA), analyses of covariance (ANCOVA), Pearson correlations, partial correlations, and multivariate regression analysis to test the influence of NRG3 rs10748842 on the aforementioned variables. All RBANS five sub-scores and total score were lower in patients than those in controls (all p < 0.001). Patients carrying NRG3 rs10748842 TC + CC heterozygous genotype had lower attention score compared to TT homozygous genotype (adjusted F = 4.77, p = 0.029). BMI was positively associated with language score in patients (beta = 0.387, t = 2.59, p = 0.01). Interestingly, we further found positive association between BMI and language score in TT carriers (partial correlations: r = 0.13, adjusted p = 0.004; multivariate regression: beta = 0.42, t = 2.66, p = 0.008), but not in CT + CC carrier (p > 0.05). Our study demonstrated that NRG3 rs10748842 was associated with cognitive impairments, especially attention performance in schizophrenia. Moreover, NRG3 rs10748842 altered the effect of BMI on cognitive impairments as measured by the RBANS language score in chronic patients with schizophrenia.