This module implements IMRPhenomD in a pure Python code, compiled with the Numba just-in-time compiler. The structure of the code is closely related to the C code; the module provides nearly identical function interfaces in IMRPhenomD.py. The module implements the analytic first and second derivatives necessary to compute t(f) and t'(f), rather than computing them numerically, as is done in the C code. Using the analytic derivatives increases the code complexity but is wall-time faster and produces more numerically accurate results. The improvement in numerical accuracy is particularly significant for t'(f). In testing, PyIMRPhenomD is considerably faster than the C implementation. For large frequency grids, the Python version's speed-up is typically approximately a factor of 5 compared to the C version.