The soil dielectric constant model reflects the relationship between the volumetric water content and the dielectric constant. The ground penetrating radar (GPR) could quickly, precisely and non-destructively obtain the water state of the soil layers. A reasonable and reliable dielectric constant model is of great significance for predicting, monitoring and exploring water migration in soils. Taking Yunnan laterite as an example, this study comprehensively considered the key physical factors (dry density, water content, temperature) of the in situ laterite dielectric properties. The dielectric properties and the influencing factors of laterite has been discussed with the laboratory tests by reshaping laterite, and the dielectric constant model was finally proposed to evaluate the natural water state of the laterite. The results show that the relative dielectric constant of laterite increases gradually with the increasing volumetric water content, dry density and temperature respectively. The water content is the most important influencing factor, secondly followed by temperature and dry density. The dielectric constant model of laterite was built based on the influence of dry density and temperature on the relationship between the relative dielectric constant and volumetric water content. The proposed multivariate model has a good prediction effect on the water content of laterite, and the prediction effects compared with other existing models are as follows: Multivariate model, Zhao model, Malicki model, Zhou model, Topp model, Liao mode and Herkelrath model. This model could be applied to evaluate the plastic limit index and the liquid limit index using the relative dielectric constant, and then finally estimate the soft and hard state grades of laterite. This study could provide a reference for the construction of laterite dielectric constant models in different regions, and a convenient way for the estimation of the laterite water content state and engineering categories.