The background that drives this research is the low numeracy ability of children in Komunitas Harapan, so it is necessary to apply the right learning model and habituate the use of appropriate applications to improve the numeracy ability. Based on the results of observations and interviews that have been carried out, 7 out of 10 children at the age of Phase C do not understand the types of story questions poured into numeracy which are currently often used and will appear in many Computer-Based National Assessment Applications (ANBK) which are used as media for the implementation of the Minimum Competency Assessment (AKM). This research method uses a quantitative approach to Pre-Experimental Design type experiments in the form of One-Group Pretest-Posttest Design. Based on the results of the research analysis, the average Posttest score of the control and experimental class was 78.75 from the results of the N-Gain Test, the score was obtained at 0.5625, which means it is included in the moderate interpretation, while the N-Gain Percent Test obtained a value of 56.25%, which means it is included in the effective category. The results of the Two Party T Test using a different test (Paired Sample t-test) obtained a calculated tcount 6.111 with a ttable of 2.365 results of a comparison tcount ? ttable means that H0 is rejected and H1 is accepted. So it can be concluded that there is an influence from the application of contextual learning models in improving children's numeracy skills in Komunitas Harapan.