This study uses a Pre-Experimental Design research type with the One Group Pretest Post-test Design model. Research subjects were selected using by non-probability sampling saturated sampling model. The samples taken were 26 students. By using data collection method in the form of a description tests of the ability to solve story problems. The validity test for the description test of the ability to solve story problems uses the Pearson formula. The reliability test used cronbach's alpha with the help of IBM SPSS 25. In addition, the researcher used a difficulty level test and a discriminatory power test. Data analysis used a non-parametric statistical technique, namely the Wilcoxon testThe results showed that the calculation of the average score of the post-test description test was 79.36 which was greater than the test results in the pre-test with an average value of 58.46. This is evidenced by the results of the Wilcoxon test analysis which shows Zcount of -4.436 with a significance value of 0.00 <0.05, meaning that there is a significant difference between the results of the pre-test and post-test using the Problem Based Learning Model with the help of fractional donut media. The results of this study can be concluded that the use of Problem Based Learning Model assisted by fractional donut media has a significant effect on the ability to solve students' story problems in Mathematics.