Artifical neural network (ANN) approaches to motor control commonly adopt generic architectures like fully connected multi-layer perceptrons (MLPs) and recurrent neural networks (RNNs). As such architectures contain few inductive biases, it is common to require priors in the form of rewards, training curricula, imitation data, or trajectory generators. In nature, animals are born with priors in the form of their nervous system's architecture, which has been shaped by evolution to confer innate ability and efficient learning. For instance, a horse can walk within hours of birth and quickly improve with practice. Can architectural priors based on neural circuits provide useful advantages for ANNs? Previous work in Neural Circuit Architectural Priors (NCAP) explored the case study of C. elegans by translating its swimming circuits into an ANN controlling a simulated nematode body. This architecture achieved good performance, data efficiency, and parameter efficiency, and its modularity facilitated interpretation and transfer to body variations. Nevertheless, it remained unknown whether such an approach could scale to more complex animals, as C. elegans nervous systems have only 302 neurons, highly stereotyped connectivity, and a mapped connectome, while mammalian nervous systems have millions or billions of neurons, more variable connectivity, and no mapped connectome. In this work, we introduce a biologically inspired ANN architecture for quadruped locomotion based on neural circuits in the limbs and spinal cord of mammals. Our architecture achieves good initial performance and comparable final performance to MLPs, while using less data and orders of magnitude fewer parameters. Our architecture also exhibits better generalization to task variations, even admitting deployment on a physical robot without standard sim-to-real methods. In ongoing work, we transfer this architecture to a simulated rodent body. Overall, our findings encourage future work in yet more complex sensorimotor skills.