Objective：To investigate the relationship between alcohol intake and prevalence of diabetes mellitus（DM）. Methods：Stratified sampling was used to select 4 000 permanent residents aged 40 years and above. All participants completed standardized questionnaire，physical examination and biochemical tests. Results：The total effective number was 3 970 and the prevalence of DM was 19.47%. The percentage of drinking was 41.2%，of which the male was 60.9%，the female was 30.9%；the percentage of male was significantly higher than that of female（P<0.05）. The prevalence of newly diagnosed DM in small-amount drinking group（male 5.09%，female 4.26%） was lower than that of non-drinking group（male 9.17%，female 7.11%）. The prevalence in large-amount drinking group（male 21.76%，female1 15.19%） was significantly higher than that of other three groups. The prevalence of previously diagnosed DM was decreased with the increase of drinking amount；the prevalence of large-amount drinking group was the lowest. In general，the prevalence of newly diagnosed DM in small-amount drinking group was lower than that of non-drinking group. The prevalence of large-amount drinking group was significantly higher than that of the other three groups. The prevalence of previously diagnosed DM in drinking groups was lower than that of non-drinking group. The levels of 2-hour postprandial glucose（2hPG） and hemoglobin A1C（HbA1C） in groups of small-amount and moderate-amount drinking groups were significantly lower than those of non-drinking group. In large-amount drinking group，the levels of fasting plasma glucose，2hPG and HbA1C were significantly higher than those of the other three groups；the level of cholesterol was significantly higher than that of small and moderate drinking group. The high-density lipoprotein was significantly lower than that of the other three groups，the low-density lipoprotein was significantly higher than that the other three groups（P<0.05）. The multivariate logistic regression analysis revealed that the group of small-amount drinking had a lower odds ratio for diabetes compared with that of non-drinking group（OR=0.641，95%CI=0.508 to 0.807），and it was a protective factor for diabetes even after the adjustment of gender，age，BMI，smoking，calorie intake，family history of diabetes. Conclusion：Small-amount drinking is a protective factor for DM. Large-amount drinking can increase the levels of blood glucose and blood lipid.