The objectives of this study are to observe whether : (1) the improvement of students’ logical reasoning ability thought by TPS cooperative learning model is higher than those taught by expository learning model, (2) the improvement of students’ mathematical communication ability taught by TPS cooperative learning model is higher than those taught by exspository learning model, This study was held at SMP Negeri 24 Medan by having 56 students as sample. This study used quasi-experimental method with pretest-postest control group design. The population of this study was all students of grade VIII taking two classes (experimental 1 class and experimental 2 class) through random sampling technique. The instrument had required content validity and coefficient reliability. Data were anayzed by t test. Before it was used t test the normality and homogenity tests with significant level 5% had been done. The result of data analysis showed that the average of N-Gain logical reasoning ability test is 0,65 in experiment class, while 0,5 in control class, with sig = 0,007 and 0.007 < α 0,05. Therefore, the improvement of students’ logical reasonig ability taught by experiment class is higher than control class.  The average of N-Gain communication ability test in experiment is 0,62 in experiment class, while and 0,38 in control class, with sig = 0,000 and 0.000 < α 0,05. Therefore, the improvement of students’ communication ability taught by experiment class is higher than control class.