In this paper we consider modeling leptokurtic daily log-return distributions of three commodities: gold, silver and crude oil. Three modeling approaches are tried out namely (a) a two component mixture of normal distributions model, (b) Variance Gamma (VG) distribution model and (c) Generalized Secant Hyperbolic (GSH) distribution model. The two component mixture of normal distributions model is found to be a reasonable model for log-returns on gold and crude oil. The VG distribution model and the GSH distribution model are found not to be suitable for modeling log-returns for any of the three commodities considered in this paper.