Diabetes mellitus is a complex chronic condition requiring daily self-management to prevent/delay the onset of diabetes-related complications. The emphasis on control for diabetes self-management can lead to feelings of self-blame and failure when targets are not reached. Self-compassion may offer an alternative way of relating to the self when such feelings arise, and in turn, positively influence diabetes health outcomes. However, little is known about how self-compassion relates to behavioural, clinical and emotional outcomes in adults with diabetes. The aim of the current study was therefore to determine the associations between self-compassion and diabetes-related health behaviours and clinical outcomes, and emotional health outcomes. Cross-sectional data from adults (N = 1907) aged 18–75 years with type 1 or type 2 diabetes, who participated in the second Diabetes MILES—Australia (MILES-2) study, were analysed. Behavioural outcomes were healthy eating and physical activity (SDSCA subscales); clinical outcome was self-reported haemoglobin A1c (HbA1c; an important measure of average blood glucose in diabetes management); emotional outcomes were depressive symptoms (PHQ-8), anxiety symptoms (GAD-7) and diabetes distress (PAID). Self-compassion was measured using the Self-Compassion Scale Short-Form (SCS-SF). Self-compassion was significantly lower among those with severe diabetes distress or moderate-to-severe symptoms of depression and anxiety, as compared to those with no/mild symptoms. Further, hierarchical linear regression analyses, split by diabetes type, revealed that self-compassion was significantly associated with all specified outcomes, with the strongest associations observed among the emotional outcomes (β range, − 0.47 to − 0.55; all p < 0.01). The findings indicate self-compassion is meaningfully associated with more optimal behavioural, clinical and, especially, emotional outcomes in adults with diabetes.