The dataset was compiled from 16 peer-reviewed literature and includes 990 mix design of geopolymer used in soil stabilization. For each mix design, the dataset provides the detailed information on soil properties, chemical composition of geopolymer, curing time, and the measured unconfined compressive strength. The compiled dataset was used to develop our machine learning model for mix design of geopolymer, which is introduced in our manuscript entitled "New Generic Framework for Mix Design of Geopolymer for Soil Stabilization: Composition-informed Machine Learning Model"