Self-efficacy is a core element of diabetes self-care and a primary target of diabetes interventions. Adults with serious mental illness (SMI) are twice as likely as adults among the general population to have Type 2 diabetes. This population faces substantial barriers (i.e., cognitive impairment, psychiatric symptoms) to optimal diabetes self-care, but the relationship of these barriers to both self-efficacy and glycemic control (hemoglobin A1C [A1C]) is not clearly understood. Data collected from adult participants with SMI (i.e., schizophrenia, schizoaffective disorder, bipolar disorder) and Type 2 diabetes (n = 92) were used to examine the moderating effects of cognitive functioning and psychiatric symptoms (i.e., positive and negative symptoms) on the association between self-efficacy and A1C. The relationship between self-efficacy and A1C was moderated by cognitive functioning (B = -4.03, standard error = 1.54, p = .011). Greater self-efficacy was associated with better glycemic control when cognitive functioning was high, but worse control when functioning was low. The relationship between self-efficacy and A1C was moderated by negative symptom severity (B = 6.88, standard error = 3.34, p = .043). Higher self-efficacy was associated with poorer glycemic control only when negative symptom severity was high. Positive symptoms did not interact with self-efficacy to predict A1C. These results suggest that adults with SMI and low cognitive function or high negative symptom severity may misperceive their ability to manage their diabetes. They may benefit from efforts, including care management and monitoring, cognitive remediation, and skill training, to identify and correct inaccurate diabetes self-efficacy.