The kinetic modelling of cyclohexane auto-ignition in a rapid compression machine is addressed, based on the comprehensive kinetic mechanism, comprising 499 species and 2323 reactions, then reduced to 56 species involved in 196 reactions and 50 species involved in 143 reactions. The purpose is to explore the merits of reduced kinetic models that can be incorporated into Computational Fluid Dynamic codes for the numerical investigation of the performance of fuels in engines, with specific reference to Homogeneous Charge Compression Ignition (HCCI) or Controlled Auto-Ignition (CAI). Calculations of ignition delay, using the SPRINT zero-dimensional code assuming adiabatic reaction, have been performed for the stoichiometric mixture of cyclohexane in air, comprising C 6H 12 + 9O 2 + 33.86N 2, at an initial pressure of 0.48 bar, yielding compressed gas pressures of 7.2-9.7 bar over the compressed gas temperature range 650-925 K. There is reasonable agreement between the predicted delays of each scheme over the whole range of compressed gas temperatures. The common feature for alkane fuels, of negative temperature dependent ignition delays in an intermediate range, is predicted. However, the agreement with experimental measurements, especially in the negative temperature dependent region, is not very satisfactory. Non-adiabatic reaction as a cause of the discrepancy is addressed, initially via simple tests using SPRINT. The reduced mechanism comprising 50 species has then been implemented into the multidimensional CFD code FLUENT, which is the maximum number of species that can be incorporated in this code. The predictions of FLUENT are in excellent agreement with those from SPRINT under closed, constant volume adiabatic conditions. When non-adiabaticity at the wall is assumed, the FLUENT calculations show how temperature gradients evolve, leading to spatial differences in the rates of development of reaction. In particular, reaction is able to evolve more rapidly in the boundary layer region at compressed gas temperatures that correspond to the region of negative temperature dependence of ignition delay, so causing a reduction in the overall ignition delay relative to that predicted under adiabatic conditions. Ignition itself is initiated in the boundary layer under these circumstances. © 2012 Elsevier Ltd. All rights reserved.