One of the fundamental questions in sampling is how to use the complete auxiliary information in the estimating stag effectively. In sample surveys, recent advanced literature on estimation using auxiliary information can be classified into three main approaches according to the methodology used in the estimating stage. These approaches are the design-based approach, model-based approach, and model-assisted approach (Wu and sitter, 2001). The design-based approach judges estimators with reference to the sampling distribution. The model-based approach used a model for the variable of interest. If the model is not satisfied, the estimator will not be efficient. The model-assisted approach estimators are approximately (asymptotically) design unbiased irrespective of whether the working model is correct or not and are particularly efficient if the working model is correct. In this approach, inferences and the asymptotic framework are designed based on the working model, which is only used to increase efficiency. Stratification is a common technique to increase the precision of the finite population estimators (Hussein, 1999), It is suggested when it is possible to divide a heterogeneous population into homogenous subpopulations. It is useful when the data of known precision are wanted or when sampling problems differ markedly in different parts of the population (Cochran, 1977). In this paper, we consider the use of a more complex model in obtaining model-assisted estimators for stratified random samples. The proposed model calibration estimators can handle any linear or nonlinear model and reduce to the conventional calibration estimators of Deville and Samdal (1992) in the linear model case.