This study investigated whether free prostate-specific antigen (fPSA) performs better than total PSA (tPSA) in predicting prostate volume (PV) in Chinese men with different PSA levels. A total of 5463 men with PSA levels of <10 ng ml-1 and without prostate cancer diagnosis were included in this study. Patients were classified into four groups: PSA <2.5 ng ml-1, 2.5-3.9 ng ml-1, 4.0-9.9 ng ml-1, and 2.5-9.9 ng ml-1. Pearson/Spearman's correlation coefficient (r) and receiver operating characteristic (ROC) curves were used to evaluate the ability of tPSA and fPSA to predict PV. The correlation coefficient between tPSA and PV in the PSA <2.5 ng ml-1 cohort (r = 0.422; P < 0.001) was markedly higher than those of the cohorts with PSA levels of 2.5-3.9 ng ml-1, 4.0-9.9 ng ml-1, and 2.5-9.9 ng ml-1 (r = 0.114, 0.167, and 0.264, respectively; all P ≤ 0.001), while fPSA levels did not differ significantly among different PSA groups. Area under ROC curve (AUC) analyses revealed that the performance of fPSA in predicting PV ≥40 ml (AUC: 0.694, 0.714, and 0.727) was better than that of tPSA (AUC = 0.545, 0.561, and 0.611) in men with PSA levels of 2.5-3.9 ng ml-1, 4.0-9.9 ng ml-1, and 2.5-9.9 ng ml-1, respectively, but not at PSA levels of <2.5 ng ml-1 (AUC: 0.713 vs 0.720). These findings suggest that the relationship between tPSA and PV may vary with PSA level and that fPSA is more powerful at predicting PV only in the ''gray zone'' (PSA levels of 2.5-9.9 ng ml-1), but its performance was similar to that of tPSA at PSA levels of <2.5 ng ml-1.