Critical Thinking Skills (CTSs) are necessary for science mastery. To help the students develop their CTSs, discovery-based learning and student-centered activities may be beneficial. This study aims at investigating: (1) effects of discovery-based multiple representation learning (DMRL) model on CTSs; (2) effects of different academic achievement on CTSs; and (3) interaction between a learning model and types of different academic achievement on CTSs. The participants were 162 students, grade 7th from 3 public schools in Sleman, Indonesia. They were classified into 54 students with exceptional academic achievement, 48 students with average academic achievement, and 60 students with below-average intellectual ability. We used an essay test to collect data on CTSs utilizing a non-equivalent control group design with pretest and posttest. ANCOVA was used to examine the data (p = .05). The results of the study show that: (1) The DMRL had a high potential to improve the students’ CTSs; (2) The students with HA had the highest CTSs; (3) there is an interaction effect between the learning model and the different types of achievement on CTS. As a result, we conclude that the DMRL was effective in closing the CTSs discrepancies between students with high, moderate, and poor academic achievement.