This study aims to compare the effectiveness of the RODE and PBL learning models in improving students’ mathematical communication skills. The study employed a quasi-experimental nonequivalent-group pretest-posttest design involving two experimental groups. The research sample consisted of 44 eighth-grade students of MTs Al-Imam Asy-Syafi’i in Tarakan, North Kalimantan, Indonesia. The students were divided into two groups: Experimental group 1 was taught using the RODE model, and experimental group 2 was taught using the PBL model. Data were collected through pretests and posttests in the form of three essay questions measuring the indicators of written text, drawing, and mathematical expression. The normality test indicated that the N-Gain data were not normally distributed; therefore, inferential analysis was conducted using the Mann–Whitney U test. The results showed that the mean N-Gain of the RODE group (0.8290) was higher than that of the PBL group (0.6107). The Mann–Whitney U test showed a significant difference between the two groups (U = 157.50, Z = -2.07, p = 0.038), with the mean rank of the RODE group (26.34) higher than that of the PBL group (18.66). Thus, there was a significant difference in the improvement of students’ mathematical communication skills between the two learning models. These findings indicate that the RODE model is more effective than the PBL model in improving students’ mathematical communication skills, particularly on the topic of linear equations in one variable at MTs Al-Imam Asy-Syafi’i in Tarakan, North Kalimantan, Indonesia.