A detailed mechanism for the four-component RD387 gasoline surrogate developed by Lawrence Livermore National Laboratory has,shown-good agreement with experiments in engine-relevant conditions. However, with 1388-species and 5933 reversible reactions, this detailed mechanism is far too large to use in practical engine simiilations. Therefore, reduction of the,detailed mechanism was performed using a multi-stage approach consisting,of the directed relation graph with error propagation and sensitivity analysis (DRGEPSA) method, unimportant reaction elimination, isomer lumping, and analytical qtasi-steady-state (QSS) reduction based on computational singular perturbation (CS]?) analysis. 4 new greedy sensitivity analysis algorithm was developed and demonstrated to be capable of:removing more species for the same error limit compared to the conventional sensitivity analysis used in directed relation, graph (DRG)-based skeletal reduction methods. Using this new greedy,algorithm, several skeletal and reduced mechanisms, were developed at varying levels of complexity and for different target condition ranges. The final skeletal and reduced mechanisms consisted of 213 and 148 species, respectively, for a lean-to-stoichiometric, low-temperature homogeneous charge compression ignition (HCCI)-like range of conditions. For a lean-to-rich, high temperature, spark ignition/compression ignition (SI/CI)-like range of conditions; skeletal and reduced mechanisms were developed With 97 and 79 species, respectively. The skeletal and reduced mechanisms in this study were produced using an error limit of 10% and validated using homogeneous autoignition simulations over engine,relevant conditions; all showed good agreement in predicting ignition delay. Furthermore, extended validation was performed, including comparison of autoignition temperature profiles, perfectly stirred reactor (PSR) temperature response Curves and extinction turning points, and laminar flame speed calculations. All of the extended validations showed results within/neat the 10% error limit, demonstrating the adequacy Of the resulting reduced chemistry.