This paper reports the development of a large-scale model of some spinal reflex circuitry, useful for studying the dynamic interactions among neuronal populations during simple behaviors. The included populations and properties of the neurons and terminals were derived from the literature, mainly on cat spinal cord. The model was conceived as a symmetrical controller of a pair of antagonistic muscles, within the behavioral domain of the stretch and Golgi-tendon-organ reflexes, and was scaled to include realistic numbers of motoneurons. Inputs to the model were fiber populations providing random synaptic drive to some of the populations and sensory stimuli appropriate for the reflexes. The resulting model contained roughly 2300 neurons in six pairs of populations. The total number of connections in the model was about 600,000, and individual postsynaptic potentials were small (0.1-0.6 mV). Model responses were calibrated by examination of their ability to reproduce known aspects of the reflexes. Published algorithms were used to construct the environment, which is easily expandable, in terms of membrane channels, neuronal geometry, and synaptic properties. The system was built to combine a system-level perspective of spinal circuitry with the single-unit perspective common in electrophysiological investigation. It provides a computational tool for system-level investigations of spinal cord similar to the tools available at the level of membrane currents.