Background Self‐compassion has been defined as the ability to be with one's feelings of suffering in a warm and caring way. Research has shown a negative association between self‐compassion and mental illness, and that low self‐compassion can make psychotherapeutic effects less likely. The ability to measure a patient's self‐compassion in a fast and reliable way is therefore important in investigating effects of psychotherapies. The aim of the present study was to evaluate the psychometric properties of the Swedish version of the Self‐Compassion Scale–Short Form (SCS‐SF) in both non‐clinical (NC) and clinical samples. Methods Cross‐sectional data were gathered in a NC community sample (n = 1,089), an eating disorder (ED) sample (n = 253) and a borderline personality disorder (BPD) sample (n = 151). All participants were asked to complete a number of questionnaires, including the SCS‐SF, and 121 participants in the NC sample repeated the assessment after 2 weeks for test–retest analysis. Results Confirmatory factor analyses supported the first‐order model suggested in previous research. Good internal consistency (Cronbach's alpha = 0.78–0.87) and test–retest reliability (intra‐class correlation = 0.84) were demonstrated for the entire scale. Results also showed good convergent validity, demonstrating moderate negative associations between self‐compassion and mental illnesses, as expected, and acceptable divergent validity, demonstrating weak positive associations between self‐compassion and quality of life and mindfulness. Discussion The correlations between the SCS‐SF and the instruments used for validation were weaker in the clinical samples than the NC sample. This may be due to difficulties measuring these constructs or that the associations differ somewhat between different populations, which could warrant further research. The results added some support to the assumption that self‐compassion may overlap with mindfulness yet still represents a distinct construct. Conclusions Analyses of the SCS‐SF provided evidence of adequate to good psychometric properties, supporting use of the scale's total sum score and a first‐order factor structure. This is in accordance with previous evaluations of the SCS‐SF, suggesting that it is a reliable and time‐efficient instrument for measuring a general level of self‐compassion. This may be important when evaluating psychotherapy and investigating self‐compassion and its influence on psychiatric illness.