Development of a robust diagnostic test for patients co-infected with human immunodeficiency virus and tuberculosis (HIV/TB) is urgently needed. We believe N6-methyladenosine (m6A)- related long non-coding RNA (lncRNAs) from the host blood could be utilized to diagnose patients co-infected with HIV/TB. In this study, differentially expressed analysis, correlation analysis, univariate logistic regression, and logistic regression with least absolute shrinkage and selection operator (LASSO) were performed in RNA-Seq dataset containing of 14 HIV/TB co-infected subjects and 11 HIV mono-infected subjects. In total, five m6A related-lncRNAs with powerful diagnostic significance for HIV/TB co-infection were identified. We then built a deep learning model based on the five m6A related-lncRNAs for accurately discriminating the HIV/TB co-infected patients from HIV mono-infected patients with an accuracy of 92.0%, a sensitivity of 92.9%, a specificity of 90.9%, and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.935. Moreover, the diagnostic performance was validated in an external cohort containing 15 HIV/TB co-infected subjects and 16 HIV mono-infected subjects of whole blood. Overall, the findings showed that deep learning model based on five m6A-related lncRNAs had a worthy diagnostic performance for HIV/TB co-infection, and these diagnostic lncRNAs associated with m6A regulator genes could play a potential role in the pathogenesis of HIV/TB co-infection.