In combustion models employing tabulated or global kinetics, the prediction of thermal NO is usually performed by either the direct resolution of Zel'dovich mechanism or the tabulation of the NO reaction rate using a laminar flamelet database and a progress variable representative of the fuel oxidation reactions, for instance temperature or a linear combination of major products mass fractions. It is known that the first method lacks accuracy if radical species such as N or O appearing in the NO reaction rate are not correctly estimated. The second method cannot lead to accurate predictions because NO reactions take place essentially when the fuel oxidation is over, therefore the NO reaction rate shows a very weak correlation with the progress variable. In this paper a new approach called NO Relaxation Approach (NORA), is proposed for the modeling of thermal NO. It allows a high accuracy when coupled with this type of combustion models. With NORA, the NO reaction rate is written as a linear relaxation towards the equilibrium value Y-NO(eq) with a characteristic time tau. Both parameters are tabulated as functions of equivalence ratio, pressure, temperature and dilution mass fraction. NORA is first validated on homogeneous internal combustion engine cases, where it closely fits the complex chemistry results. It is then integrated into the turbulent combustion model ECFM3Z dedicated to piston engine applications. In this model a mixed tabulated (TKI) and global kinetics (CORK) approach is used to describe turbulent combustion. First applications on eight Diesel engine operating points show a good improvement with NORA compared to a classical resolution of Zel'dovich mechanism. (C) 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.