Blood glucose control is important for Type I diabetes patient. In this work we have proposed multi model based nonlinear controller design for Type I diabetic system. The optimal numbers of local models are identified by using standard gap metric technique. For each local model the linear PI controllers are designed using standard tuning procedure. The controller output are combined using T-S fuzzy and crisp technique in order to obtain the global controller output. The performances of controllers are tested with servo and regulatory cases. © 2009 Crown Copyright.