Objective: Immune checkpoint inhibitors have revolutionized cancer therapy for multiple types of solid tumors, but as expected, a large percentage of patients do not show durable responses. Biomarkers that can predict clinical responses to immunotherapies at diagnosis are therefore urgently needed. Herein, we determined the associations between baseline gut commensal microbes and the clinical treatment efficiencies of patients with thoracic neoplasms during anti-programmed death protein 1 (PD-1) therapy.Methods: Forty-two patients with advanced thoracic carcinoma who received anti-PD-1 treatment were enrolled in the study. Baseline and time-serial stool samples were analyzed using 16S ribosomal RNA gene sequencing. Tumor responses, patient progression-free survival, and overall survival were used to measure clinical outcomes.Results: The diversities of the baseline gut microbiota were similar between responders (n = 23) and nonresponders (n = 19). The relative abundances of the Akkermansiaceae, Enterococcaceae, Enterobacteriaceae, Carnobacteriaceae and Clostridiales Family XI bacterial families were significantly higher in the responder group. These 5 bacterial families acted as a commensal consortium and better stratified patients according to clinical responses (P = 0.014). Patients with a higher abundance of commensal microbes had prolonged PFS (P = 0.00016). Using multivariable analysis, the abundance of the commensal consortium was identified as an independent predictor of anti-PD-1 immunotherapy in thoracic neoplasms (hazard ratio: 0.17; 95％ confidence interval: 0.05–0.55; P = 0.003).Conclusions: Baseline gut microbiota may have a critical impact on anti-PD-1 treatment in thoracic neoplasms. The abundance of gut commensal microbes at diagnosis might be useful for the early prediction of anti-PD-1 immunotherapy responses.