Objective To investigate the relationship between anxiety sensitivity (AS), resilience and mental health of soldiers. Methods A total of 260 soldiers were selected from an army troop, and were surveyed by anxiety sensitivity index Ⅲ (ASI-3), Conner-Davidson resilience scale (CD-RISC) and self-reporting inventory (SCL-90). The relationships of these variables were analyzed by structure equation modeling analysis. Results Totally 240 (92.3%) valid questionnaires were collected. There were no significant differences in AS or mental health between soldiers with different demographic factors including service length, single-child or not, job division, educational level, or residence before entering army (P>0.05). The total score of SCL-90 and each factor scores of soldiers with high-AS were significantly higher than those with low-AS (P<0.01). The correlation analysis results showed that the total score of SCL-90 was positively correlated with the total score and each factor score of ASI-3 (r_ASI=0.487, r_somatic=0.435, r_social=0.455, r_cognitive=0.445; P<0.001), and was negatively correlated with each factor score of CD-RISC (r_tenacity=0.315, r_strength=0.321, r_optimism=0.227; P<0.001). Structural equation modeling analysis and Sobel test showed that AS indirectly affected the mental health of soldiers through the mediating effect of resilience (χ²/df=20.222/12=1.685 2, P=0.063, goodness of fit index[GFI]=0.977, adjusted goodness of fit index[AGFI]=0.946, normed fit index[NFI]=0.977, relative fit index[RFI]=0.960, incremental fit index[IFI]=0.991, Tucker-Lewis index[TLI]=0.983, comparative fit index[CFI]=0.990, and root mean square error approximation[RMSEA]=0.054). Conclusion RS has partial mediating effect on the relationship between AS and mental health of soldiers. More attention should be paid on the mental health of soldiers with high AS, so as to improve their RS and mental health.