The GISS global climate model (Hansen et al.) has been extended to include the middle atmosphere up to an altitude of approximately 85 km. The model has the full array of processes used for climate research, i.e., numerical solutions of the primitive equations, calculation of radiative and surface fluxes, a complete hydrologic cycle with convective and cloud cover parameterizations, etc. In addition, a parameterized gravity wave drag formulation has been incorporated, in which gravity-wave momentum fluxes due to flow over topography, wind shear and convection are calculated at each grid box, using theoretical relationships between the grid-scale variables and expected source strengths. The parameterized waves then propagate vertically upward depending on the instantaneous wind and temperature profiles, with waves breaking at levels in which their momentum flux exceed the background saturation value. Radiative damping is also calculated, and the total momentum convergence in each layer is used to alter the local wind, while the kinetic energy dissipation warms the temperature. Thus the generation, propagation, breaking and drag are all a function of the calculated variables at each grid box for the various vertical levels. The model has been run for five years, and the results compared with observations. The model produces generally realistic fields of temperature and wind throughout the atmosphere up to approximately 75 km. Important aspects of the current simulation include a proper break between the tropospheric and stratospheric jets, realistic closing off of the wintertime jet in the mesosphere, the observed warm winter/cold summer mesosphere, and a semiannual wind oscillation near the stratopause. The most obvious deficiencies are that the long-wave energy itself is somewhat too small in the low and midstratosphere, temperatures are too cold near the model top and are too warm in the polar Southern Hemisphere lower stratosphere during winter. Also, the model generates an inertial oscillation near the equatorial stratopause which may be excessive. Experiments are run without the various gravity wave drag mechanisms to quantify their effects. It is shown that a coarse-grid general circulation model with parameterized gravity-wave drag can produce a reasonable simulation of the middle atmosphere, which makes possible relatively long-term integrations.