Soil Water Retention Curve (SWRC) is a fundamental relationship in unsaturated soil mechanics, knowledge of which is essential for determining major mechanical and hydraulic properties of unsaturated soils. There are several empirical, semi-empirical and physically-based models which have been proposed to date for estimating SWRC. While the physically-based models which employ the basic soil characteristics such as grain-size and pore-size distributions are regarded superior to purely empirical models, their Achilles’ heel is the several simplifying assumptions based on which these models are constructed, thereby, restricting their applications and influencing their accuracy. Given the complexity of the soil porous structure, one may resort to the new inference techniques rather than mechanistic modelling to find the relationship among soil physical characteristics and the retention properties. Therefore, an alternative approach to purely empirical relationships as well as physically-based and conceptual models for determining SWRC is the use of Artificial Intelligence (AI) based techniques to acquire a relationship for SWRC based on the soil basic properties, especially grain size distribution and porosity. Among AI-based methods, Multi-Gene Genetic Programming (MGGP), often used to establish a close form equation for a complex physical system, offers a suitable alternative to the current approaches. In this study, a database compromising of 437 soils (containing various soil types, namely, sand, clay, silt, loam, silt loam, clay loam, sandy loam, sandy clay loam, silty clay loam, silty clay, and loamy sand soils) was used along with MGGP to establish a relationship among suction, saturation, porosity and grain size distribution. The proposed equation had a reasonable agreement with the experimental data compared to the other grain-based and physically-based models.