The purpose of this research is to examine measurement invariance of collaborative problem solving skills measured by PISA 2015 Xandar subtest for Singapore, Norway, and Turkey. The research was conducted with 2990 participants' data obtained from Turkey (1032), Norway (923), and Singapore (1035) on PISA 2015 collaborative problem solving study. In the first part of the study, exploratory factor analysis was performed to obtain the factor structure of the Xandar subtest. Then, the model data fit was checked by confirmatory factor analysis via chi(2)/ df(3.127), RMSEA (0.027), CFI (0.987) and TLI (0.979) values. Multi-group confirmatory factor analysis was used in invariance analyses. The findings show that the collaborative problem solving model met only configural invariance across the countries and has not met the metric, scale, and strict invariance stages. The results show that meaningful comparisons cannot be made between the countries, because the factor loadings, variances, error variances, and covariances differ among countries.