In parallel computing, 'host and device code' refers to the distinction between code that runs on the CPU (host) and code that runs on the GPU or other accelerators (device). This separation allows for the efficient execution of tasks by leveraging the strengths of both processing units, with the host managing the overall program flow and the device performing computationally intensive tasks.