This study aims to describe the ability of mathematical connections using cooperative models oftypes connecting, organizing, reflecting and extending (CORE) better than conventional learning. This study uses a quantitative approach with the experimental method. The study design was a post test only control design with a population of class IX students totaling 5 classes. The sampling technique uses simple random sampling. The selected sample is class IXC and class IXE. Data collection techniques use tests in the form of essay questions. Analysis techniques for testing hypotheses with t-test, because the data are normally distributed and homogeneous variance. From the data analysis it can be concluded that the mathematical connection ability using cooperative models oftypes connecting, organizing, reflecting and extending (CORE) is better than using conventional learning models of class IX students at Merangin 2 Public High School.