This research aims to examine and determine the improvement in problem-solving abilities with the implementation of the PBL learning model assisted by LKPD. This study was conducted at SMP Deli Murni Sukamaju in Class VIII, with sample VIII-1 as the control class and VIII-2 as the experimental class. The type of research used is quasi-experimental research. The instrument used for data collection consists of 7 pre-test questions and 7 post-test questions on problem-solving abilities in the topic of number patterns. Based on the results of the research hypothesis using a two-way ANOVA with N-Gain data, the significant value is > =0.05, so H0 is rejected both overall and based on the students' Initial Mathematical Ability. This means that the improvement in students' problem-solving abilities using the PBL model assisted by LKPD is better overall and based on the students' Initial Mathematical Ability (KAM) than students using the Conventional model. The absence of interaction between learning and KAM, as seen in the significant value of the KAM*Class row with a result of >=0.05, means that the improvement in problem-solving ability occurs entirely due to learning. Continued with the Scheffe test, which showed that the group with low KAM significantly had better improvement compared to the students in the medium and high KAM groups.