Simulating a high-resolution multimillion cell model brings many benefits, by enabling reservoir engineers to use the best grid size for accurate representation of water and gas movement in the reservoir, essential for advanced field management, Enhanced Oil Recovery or complex well design studies. To improve the characterization of a giant heterogeneous carbonate reservoir and enhance the quality of field development plans, new high-resolution static and dynamic models have been used to study one of the largest oil fields in Abu Dhabi. A detailed static model of over 50 million grid cells was constructed, using a unique water saturation modeling approach, without upscaling to a dynamic simulation, using hysteresis for both relative permeability and capillary pressure. The reservoir has over 50 years of history, with hundreds of vertical and horizontal wells. Large volumes of data from well logs, cores and other measurements were used to populate the static model, define dynamic rock types and match well log water saturation and water capillary pressure profiles. The concept of wettability change with depth was introduced, with an oil-wet system at the crest, graduating to a water-wet system near the thin transition zone. A geological resolution grid was used for reservoir simulation studies, after testing input data consistency and stable behavior. A stability test was performed by running the simulation with no wells for 50 years after equilibration and showed no movable fluids. This verified the consistency of the reservoir static properties, rock types, water saturation, relative permeability and fluid model. A history matched case was developed with over 850 wells using the same fine grid, to meet the objective of completing each simulation run within one day. After history matching, a compositional simulation model was built, to investigate the impact of grid resolution on future production forecasts. This is the largest dynamic model built by the company and demonstrates the benefits of rigorous attention to the quality of the static data, while using modern simulation workflows to avoid compromising the detailed model by upscaling. The methodologies presented in this paper will be adopted as best practices for future similar projects.