This paper presents a new coordination method of multiple model predictive controllers (MPCs) for the chemical process. In this control scheme, multiple MPCs are implemented in the control system and they run in parallel. The input to the controlled process is calculated by an affine combination of these MPC outputs. Each coordination function is made by using the soft-max function. We applied this method to temperature control for a continuous stirred tank reactor (CSTR). The simulation result showed that the proposed method provides better control performance than individual MPCs. We also compared this method with an MPC with a scheduling rule of the Q-R weights.