A phenomenological combustion model for HCCI applications - the 3-Stage heat release model - which predicts low, intermediate and high temperature heat release (LTR, ITR and HTR) is developed and validated for six automotive surrogate fuels. The contribution of all three combustion phases to the overall heat release and their durations are modeled based on rapid compression and expansion (RCEM) single-zone adiabatic simulations with detailed chemical kinetic mechanisms for all six fuels. This allows "pure" chemical effects on combustion to be included in the proposed 3-Stage heat release model, avoiding any combustion chamber related uncertainties such as blow-by, heat losses or trapped mass in crevices. Agreement between the 3-Stage heat release model and virtual RCEM (VRCEM) reference data is good. In a second step, a zero-dimensional HCCI model (0D model) is presented. To predict the onset of combustion, the low temperature ignition term of a previously developed and validated 3-Arrhenius auto-ignition model is employed in conjunction with the ignition integral of Livengood and Wu. Thereafter the 3-Stage heat release model predicts chemical heat release. Results show that the 0D model gives excellent predictions of ignition delays and combustion progress compared to the single-zone RCEM simulations using detailed chemical reaction mechanisms. Furthermore, including ITR in the 0D model is shown to be of essential importance to correctly predict high temperature ignition delay. As a final step a 3D HCCI model (3D model) is presented. The onset of combustion is again predicted by the low temperature ignition term of the 3-Arrhenius auto-ignition model, implemented via the source term of the transport equation into the 3D-CFD code. To predict heat release, the source term of the enthalpy equation is used to incorporate the 3-Stage heat release model into the 3D-CFD code. A simplified RCEM geometry is used for the simulations. Validation of the 3D model is made by a comparison to a chemical kinetic mechanism that is directly integrated into the 3D-CFD code. Combustion progress, temperature stratification in the combustion chamber and ignition delay times agree well with the reference data. The combination of the 3-Arrhenius auto-ignition model and 3-stage heat release model, their integration into 0D or 3D-CFD as a fully predictive HCCI model and their very low computational cost suggest the proposed model constitutes a promising approach for the development of future HCCI combustion engines.