(1) Background: The cognitive reserve (CR) concept has not been precisely defined in severe mental disorders and has been estimated using heterogeneous methods. This study aims to investigate and develop the psychometric properties of the Cognitive Reserve Assessment Scale in Health (CRASH), an instrument designed to measure CR in people with severe mental illness; (2) Methods: 100 patients with severe mental illness (non-affective psychoses and affective disorders) and 66 healthy controls were included. The internal consistency and convergent validity of CRASH were assessed. Spearman's correlations coefficients were also performed to examine the relationship between CRASH and neuropsychological tests, psychosocial functioning, and clinical course; (3) Results: The internal consistency was high (Cronbach's alpha coefficient = 0.903). The CRASH global score had a large positive correlation with the Cognitive reserve questionnaire total score (r = 0.838, p < 0.001), demonstrating good convergent validity. The correlation coefficients between the CRASH total scores and clinical, functional, and neuropsychological performance were different between groups. In order to provide clinical interpretation, severity classification based on diagnosis (non-affective psychotic disorders, affective disorders, and healthy controls) have been created; (4) Conclusions: CRASH is the first CR measure developed specifically for patients with severe mental illness, facilitating reliable and valid measurement of this construct. The scale may aid in the stratification of patients and the implementation of personalized interventions.