A compartmental model for diabetes is developed. The model describes the dynamics of the spread of Type 2 diabetes. A theoretical analysis in the non-adherence to drugs is investigated. A system of differential equations is analyzed by stability analysis; the non-trivial critical point obtained is locally asymptotically firm under the given conditions. In consideration of Mathematical model for glucose tolerance test (GTT) is considered, actual glucose data values are fixed using MATLAB least squares curve fitting technique. Two methods are used to numerically work out the distributions of steady states of diabetic sub-populations. The Gauss-Seidel method is more accurate than the Jacobi method. The result show that more than 50% of clinical diagnosis attempt needs to be applied to have more diagnosed population than undiagnosed. The GTT model shows that if severe diet and medication is followed diabetes can be controlled.