Hybrid multizonal/computational fluid dynamics (CFD) models provide a means of introducing a more realistic description of fluid mechanics and mixing phenomena within process models. The solution of the CFD submodel implicitly defines a function relating one or more of the variables in the multizonal model (the function outputs) in terms of another subset of the variables (the function inputs). This paper is concerned with the accurate and efficient evaluation of this function using local approximate models (LAMS) that may have a general mathematical structure, or be based on physical correlations. Practical issues relating to the robustness of the solution of hybrid models are also considered, and a general architecture for the software interface between the multizonal and CFD submodels is presented. The effectiveness and efficiency of the overall approach are tested by two applications relating, respectively, to a stirred-tank chemical reactor fitted with a. cooling jacket and to a stirred-tank bioreactor. (c) 2005 American Institute of Chemical Engineers