A previously described methodology for deriving a reduced kinetic mechanism for alkane oxidation and tested for n-heptane is here shown to be valid, in a slightly modified version, for iso-octane and its mixtures with n-pentane, iso-hexane and n-heptane. The model is still based on partitioning the species into lights, defined as those having a carbon number smaller than 3, and heavies, which are the complement in the species ensemble, and mathematically decomposing the heavy species into constituents which are radicals. For the same similarity variable found from examining the n-heptane LLNL mechanism in conjunction with CHEMKIN II, the appropriately scaled total constituent molar density still exhibits a self-similar behavior over a very wide range of equivalence ratios, initial pressures and initial temperatures in the cold ignition regime. When extended to larger initial temperatures than for cold ignition, the self-similar behavior becomes initial temperature dependent, which indicates that rather than using functional fits for the enthalpy generation due to the heavy species' oxidation, an ideal model based on tabular information extracted from the complete LLNL kinetics should be used instead. Similarly to n-heptane, the oxygen and water molar densities are shown to display a quasi-linear behavior with respect to the similarity variable, but here their slope variation is no longer fitted and instead, their rate equations are used with the ideal model to calculate them. As in the original model, the light species ensemble is partitioned into quasi-steady and unsteady species; the quasi-steady light species mole fractions are computed using the ideal model and the unsteady species are calculated as progress variables using rates extracted from the ideal model. Results are presented comparing the performance of the model with that of the LLNL mechanism using CHEMKIN II The model reproduces excellently the temperature and species evolution versus time or versus the similarity variable, with the exception of very rich mixtures, where the predictions are still very good but the multivalued aspect of these functions at the end of oxidation is not captured in the reduction. The ignition time is predicted within percentages of the LLNL values over a wide range of equivalence ratios, initial pressures and initial temperatures. (C) 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.