Results are presented from a high-resolution global ocean model that is driven through three decadal cycles of increasingly realistic prescribed atmospheric forcing from the period 1985--1995. The model used (the Parallel Ocean Program) is a z level primitive equation model with active thermohaline dynamics based on the formulation of Bryan <1969> rewritten for massively parallel computers. Improvements to the model include an implicit free-surface formulation of the barotropic mode <Dukowicz and Smith, 1994> and the use of pressure averaging for increasing the numerical time step. This study extends earlier 0.5¿ simulations of Semtner and Chervin <1992> to higher horizontal resolution with improved treatments of ocean geometry and surface forcing. The computational grid is a Mercator projection covering the global ocean from 77 ¿N to 77 ¿S and has 20 vertical levels. Three successive simulations have been performed on the CM-5 Connection Machine system at Los Alamos using forcing fields from the European Centre for Medium-Range Weather Forecasts (ECMWF). The first run uses monthly wind stresses for 1985--1995 and restoring of surface temperature and salinity to the Levitus <1982> seasonal climatology. The second run is the same but with 3 day--averaged rather than monthly averaged wind stress fields, and the third is the same as the second but uses the monthly climatological ECMWF heat fluxes of Barnier et al. <1995> instead of restoring to climatological sea surface temperatures. Many features of the wind-driven circulation are well represented in the model solutions, such as the overall current patterns, the numerous regions of hydrodynamic instability which correspond to those observed by satellite altimetry, and the filamented structure of the Antarctic Circumpolar Current. However, some features such as the separation points of the Gulf Stream and Kuroshio and the transport through narrow passages such as the Florida Straits are clearly inaccurate and indicate that still higher resolution may be required to correct these deficiencies. Water mass properties and some aspects of the thermohaline circulation are also not always well reproduced, which is partly due to the relatively short length of the integrations. The use of the ECMWF heat fluxes, rather than restoring to climatological surface temperatures, leads to stronger and more realistic surface and deep western boundary currents (primarily in the Atlantic) as well as more realistic meridional heat transport; this is primarily because the equilibrium meridional heat transport implied by the ECMWF surface fluxes is quite large. The ECMWF heat fluxes also produce improved seasonal cycles of sea surface temperature and height inboth the northern and southern hemispheres. The 3-day wind forcing gives rise to modes of model variability that are clearly seen in synoptic observations, such as the large-scale 20--100-day oscillations seen in the TOPEX/POSEIDON data, which are barotropic oscillations induced by the high-frequency wind forcing. Additional studies on other aspects of the simulations described here are being conducted by a variety of investigators, and some of these are briefly described.