Radiotherapy is commonly used to treat cancer, and localized energy deposited by radiotherapy has the potential to chemically uncage prodrugs; however, it has been challenging to demonstrate prodrug activation that is both sustained in vivo and truly localized to tumors without affecting off-target tissues. To address this, we developed a series of novel phenyl-azide-caged, radiation-activated chemotherapy drug-conjugates alongside a computational framework for understanding corresponding pharmacokinetic and pharmacodynamic (PK/PD) behaviors. We especially focused on an albumin-bound prodrug of monomethyl auristatin E (MMAE) and found it blocked tumor growth in mice, delivered a 130-fold greater amount of activated drug to irradiated tumor versus unirradiated tissue, was 7.5-fold more efficient than a non albumin-bound prodrug, and showed no appreciable toxicity compared to free or cathepsin-activatable drugs. These data guided computational modeling of drug action, which indicated that extended pharmacokinetics can improve localized and cumulative drug activation, especially for payloads with low vascular permeability and diffusivity and particularly in patients receiving daily treatments of conventional radiotherapy for weeks. This work thus offers a quantitative PK/PD framework and proof-of-principle experimental demonstration of how extending prodrug circulation can improve its localized activity in vivo.