High Pressure Air Injection (HPAI) is a potentially attractive enhanced oil recovery method for deep, high-pressure light oil reservoirs after waterflooding. The advantage of air over other injectants, like hydrocarbon gas, carbon dioxide, nitrogen, or flue gas, is its availability at any location. HPAI has been successfully applied in the Williston Basin for more than twenty years and is currently being considered by many operators for application in their assets. Evaluation of the applicability of HPAI requires conducting laboratory experiments under reservoir temperature and pressure conditions to confirm crude auto-ignition and to assess the burn characteristics of the crude/reservoir rock system. The ensuing estimation of the potential incremental recovery from the application of HPAI in the reservoir under consideration requires fit-for-purpose numerical modeling. Typically, the flue gas generated in-situ by combustion leads to in an immiscible gas drive, where the stripping of volatile components is a key recovery mechanism. HPAI has therefore, in some instances, been modeled as an isothermal flue gas drive, employing an Equation of State (EOS) methodology. This approach, however, neglects combustion and its effects on both displacement and sweep. Furthermore, the EOS approach cannot predict if, and when, oxygen breakthrough at producers occurs. Combustion can be included in a limited fashion in simulations at the expense of extra computational time and complexity. In the available literature, combustion is taken generally into account under quite simplified conditions. This paper addresses the role that combustion plays on the incremental recovery of HPAI. Numerical simulations were conducted in a 3D model with real geological features. In order to capture more realistically the physics of the combustion front, a reservoir simulator with dynamic gridding capabilities was used. Kinetic parameters were based on the combustion tube laboratory experiments. The impact of combustion on residual oil, sweep efficiency and predicted project lifetime is presented by comparing isothermal EOS-simulations and multi-component combustion runs.