In this study, by considering the changes in inlet pressures, connection nozzle materials and the number of orifices, the performances of Ranque–Hilsch vortex tubes (RHVTs) were modelled by multiple linear regression method at which carbon dioxide (CO2) was used as compressed fluid, and optimum operating conditions were tried to be determined by Taguchi method. In the experiment, counterflow RHVTs having O10 mm inner diameter and 100 mm body length were used, hot fluid outlet of which was left completely open. In the tests designed using Taguchi L16 mix type (8 × 1 2 × 3) orthogonal array, 8 different inlet pressures (200, 250, 300, 350, 400, 450, 500 and 550 kPa), 2 different types of connections (parallel and series), 2 different types of nozzle materials (brass and polyamide) and 2 different numbers of orifices (4 and 6) were chosen as control factors and levels. Maximum temperature gradient (∆T) was determined as quality characteristics, and RHVT’s performance tests were done. In order to determine the effect rates of selected control factors on quality characteristics (∆T), analysis of variance (ANOVA) was applied within 95% confidence level. According to the statistical results, it was determined that ΔT values were mostly influenced by the inlet pressure and this was followed by connection type, number of orifices and nozzle material. With the confirmation test performed, reliability of the optimization was tested and confirmed.