The aim of the present study was to explore the influence of tea consumption on diabetes mellitus in the Chinese population. This multi-centre, cross-sectional study was conducted in eight sites from south, east, north, west and middle regions in China by enrolling 12 017 subjects aged 20–70 years. Socio-demographic and general information was collected by a standardised questionnaire. A standard procedure was used to measure anthropometric characteristics and to obtain blood samples. The diagnosis of diabetes was determined using a standard 75-g oral glucose tolerance test. In the final analysis, 10 825 participants were included and multiple logistic models and interaction effect analysis were applied for assessing the association between tea drinking with diabetes. Compared with non-tea drinkers, the multivariable-adjusted OR for newly diagnosed diabetes were 0·80 (95 % CI 0·67, 0·97), 0·88 (95 % CI 0·71, 1·09) and 0·86 (95 % CI 0·67, 1·11) for daily tea drinkers, occasional tea drinkers and seldom tea drinkers, respectively. Furthermore, drinking tea daily was related to decreased risk of diabetes in females by 32 %, elderly (>45 years) by 24 % and obese (BMI > 30 kg/m2) by 34 %. Moreover, drinking dark tea was associated with reduced risk of diabetes by 45 % (OR 0·55; 95 % CI 0·42, 0·72; P < 0·01). The results imply that drinking tea daily was negatively related to risk of diabetes in female, elderly and obese people. In addition, drinking dark tea was associated with decreased risk of type 2 diabetes mellitus.