Aim To validate the German Diabetes Risk Score within the population‐based cohort of the Cardiovascular Disease – Living and Ageing in Halle (CARLA) study. Methods The sample included 582 women and 719 men, aged 45–83 years, who did not have diabetes at baseline. The individual risk of every participant was calculated using the German Diabetes Risk Score, which was modified for 4 years of follow‐up. Predicted probabilities and observed outcomes were compared using Hosmer–Lemeshow goodness‐of‐fit tests and receiver–operator characteristic analyses. Changes in prediction power were investigated by expanding the German Diabetes Risk Score to include metabolic variables and by subgroup analyses. Results We found 58 cases of incident diabetes. The median 4‐year probability of developing diabetes based on the German Diabetes Risk Score was 6.5%. The observed and predicted probabilities of developing diabetes were similar, although estimation was imprecise owing to the small number of cases, and the Hosmer–Lemeshow test returned a poor correlation (chi‐squared = 55.3; P = 5.8*10‐12). The area under the receiver–operator characteristic curve (AUC) was 0.70 (95% CI 0.64–0.77), and after excluding participants ≥66 years old, the AUC increased to 0.77 (95% CI 0.70–0.84). Consideration of glycaemic diagnostic variables, in addition to self‐reported diabetes, reduced the AUC to 0.65 (95% CI 0.58–0.71). A new model that included the German Diabetes Risk Score and blood glucose concentration (AUC 0.81; 95% CI 0.76–0.86) or HbA1c concentration (AUC 0.84; 95% CI 0.80–0.91) was found to peform better. Conclusions Application of the German Diabetes Risk Score in the CARLA cohort did not reproduce the findings in the European Prospective Investigation into Cancer and Nutrition (EPIC) Potsdam study, which may be explained by cohort differences and model overfit in the latter; however, a high score does provide an indication of increased risk of diabetes. What's new? The German Diabetes Risk Score (DRS), developed within the European Prospective Investigation into Cancer and Nutrition (EPIC) Potsdam study, has not been prospectively validated in a representative sample of the general population apart from EPIC. The aim of this study was to validate the DRS in the Cardiovascular Disease – Living and Ageing in Halle (CARLA) study, a population‐based cohort with a high response rate, that provides a more representative sample than the EPIC‐Potsdam study. The predictive ability of the DRS observed in the CARLA cohort was consistently lower than in the EPIC cohort, a finding that could reflect cohort differences and model overfit in the EPIC study.