Introduction Both body mass index (BMI) and waist circumference (WC) are associated with diabetes risk, and the difference between them in predictive ability for diabetes is still contentious. We conducted a population-based study to investigate and compare the association of them with diabetes by sex. Methods This study included a total of 4754 subjects aged 40–80 years with no diabetes at baseline between 2008 and 2017. Using multivariate Cox proportional hazards models, we calculated hazard ratios for diabetes according to tertiles of BMI or WC. Harrell's C statistics was applied to assess and compare the predictive ability of the models using BMI and WC. Results Both BMI and WC showed the significant positive trends with diabetes risk. In men, the extreme tertiles (BMI > 25.1 kg/m 2 and WC > 88.0 cm) provided 1.58-fold or 2.04-fold higher risk compared with the first tertiles (< 22.6 kg/m 2 and < 81.2 cm). In women, BMI > 24.4 kg/m 2 showed 3.28-fold higher risk than the first tertile (< 21.6 kg/m 2 ), whereas WC ≥ 78.2 cm was more than twice as likely to suffer from diabetes as WC < 78.2 cm. BMI and WC showed a comparative performance in predicting diabetes in both sexes ( P value 0.447 in men, and 0.337 in women). Conclusion Both BMI and WC showed a positive association with diabetes and offered a comparative predictive performance for diabetes in both sexes. The cut-off points, BMI 25.1 kg/m 2 and WC 88.0 cm in men and BMI 24.4 kg/m 2 and WC 78.2 cm in women, might contribute to the effective prevention strategies for diabetes.