We aimed to verify by Rasch analysis whether the Mini-BESTest, a balance measure, confirms its main psychometric properties in patients with subacute stroke undergoing rehabilitation in three different countries (Slovenia, Croatia, and Italy), and to examine the stability of item hierarchy and difficulty across the three national versions through a differential item functioning analysis. We investigated 159 patients with subacute stroke consecutively admitted to three rehabilitation facilities after screening for an intensive, tailored rehabilitation program. Balance function was tested within 36 h from admission and after ∼25 days. As no differential item functioning was found between admission and discharge data or among countries, all data were pooled. Rasch criteria for the functioning of rating scale categories were fulfilled. In terms of internal construct validity, all items except item #14 (Cognitive Get Up & Go; infit value=1.42) showed an acceptable fit to the Rasch model. The patient ability–item difficulty matching was very good. Reliability indices were high. The Principal Component Analysis of standardized residuals confirmed the unidimensionality of the test. On the basis of the item calibration, raw scores of the Mini-BESTest were transformed into linear estimates of dynamic balance and six statistically detectable levels of balance ability were defined. Good psychometric features of the Mini-BESTest were confirmed. The three different national versions showed stability in item hierarchy, indicating equivalence of their cross-cultural adaptations. Problems with item #14 in these patients warrant further study.