This article discusses ways in which earthquake rupture forecast models might be improved. Because changes are most easily described in the context of specific models, the third Uniform California Earthquake Rupture Forecast (UCERF3) and its presumed successor, UCERF4, is used as a basis for discussion. Virtually all of the issues and possible improvements discussed are nevertheless general and should therefore be applicable to other regions as well. Two common themes are a need for better epistemic uncertainty representation and the potential utility of physics‐based simulators. Given the large number of possible improvements, coupled with challenges in defining the potential value of each, which will vary among uses, community feedback is invaluable in terms of setting priorities. We should also strive to define more objective valuation metrics.