Abstract Background Low literacy and numeracy skills are common. Adequate numeracy skills are crucial in the management of diabetes. Diabetes patients use numeracy skills to interpret glucose meters, administer medications, follow dietary guidelines and other tasks. Existing literacy scales may not be adequate to assess numeracy skills. This paper describes the development and psychometric properties of the Diabetes Numeracy Test (DNT), the first scale to specifically measure numeracy skills used in diabetes. Methods The items of the DNT were developed by an expert panel and refined using cognitive response interviews with potential respondents. The final version of the DNT (43 items) and other relevant measures were administered to a convenience sample of 398 patients with diabetes. Internal reliability was determined by the Kuder-Richardson coefficient (KR-20). An a priori hypothetical model was developed to determine construct validity. A shortened 15-item version, the DNT15, was created through split sample analysis. Results The DNT had excellent internal reliability (KR-20 = 0.95). The DNT was significantly correlated (p < 0.05) with education, income, literacy and math skills, and diabetes knowledge, supporting excellent construct validity. The mean score on the DNT was 61% and took an average of 33 minutes to complete. The DNT15 also had good internal reliability (KR-20 = 0.90 and 0.89). In split sample analysis, correlations of the DNT-15 with the full DNT in both sub-samples was high (rho = 0.96 and 0.97, respectively). Conclusion The DNT is a reliable and valid measure of diabetes related numeracy skills. An equally adequate but more time-efficient version of the DNT, the DNT15, can be used for research and clinical purposes to evaluate diabetes related numeracy.