We propose a model based bolus calculator which, based on an identified model, information on the carbohydrate content of the meals and an initial single glucose measurement, computes the optimal insulin bolus dose in order to minimize the blood glucose deviations. We use a simulation model for identification of models for control and closed-loop simulations. The performance of the proposed bolus calculator is compared with a model predictive control approach which uses continuous glucose information and insulin injection. Both methods achieve comparable performance for some patients while the insulin amount, number of injections and computational complexity is significantly reduced for the bolus calculator. © 2014 American Automatic Control Council.