We present new deepCR models for use on all filters of Hubble ACS/WFC. deepCR is a deep-learning-based cosmic ray rejection algorithm previously demonstrated to be superior to state-of-the-art LACosmic for the F606W filter on Hubble ACS/WFC. We construct new training sets from F435W, F606W, and F814W filters with three types of fields: globular clusters, extragalactic fields, and resolved galaxies. We build both separate models for each filter and all-in-one global model, which can be applied to any ACS/WFC filter as the training sets cover the full span of the ACS/WFC filter range. We demonstrate near 100% detection rates of CRs in extragalactic fields and globular clusters and 91% in resolved galaxy fields.