Calcitonin gene-related peptide (CGRP) signaling inhibitors have shown efficacy in both the acute and preventive treatment of migraine. Telcagepant, a first-generation CGRP receptor antagonist, was effective but failed in clinical trials due to hepatotoxicity. Subsequently, while four next-generation CGRP receptor antagonists (rimegepant, zavegepant, atogepant, and ubrogepant) were being advanced into late-stage clinical trials, due to telcagepant’s failure, more confidence in the liver safety of these compounds was needed. DILIsym v6A, a quantitative systems toxicology (QST) model of drug-induced liver injury (DILI), was used to model all five compounds and thus to compare the four next-generation CGRP receptor antagonists to telcagepant. In vitro experiments were performed to measure the potential for each compound to inhibit bile acid transporters, produce oxidative stress, and cause mitochondrial dysfunction. Physiologically-based pharmacokinetic (PBPK) models were produced for each compound in order to appropriately estimate liver exposure. DILIsym predicted clinical elevations of liver enzymes and bilirubin for telcagepant, correctly predicting the observed DILI liability of the first-generation compound. By contrast, DILIsym predicted that each of the four next-generation compounds would be significantly less likely to cause DILI than telcagepant. Subsequent clinical trials have validated these predictions for each of the four compounds, and all 3 of the compounds submitted to FDA to date (rimegepant, ubrogepant and atogepant) have since been approved by the FDA with no warning for hepatotoxicity. This work demonstrates the potential for QST modeling to prospectively differentiate between hepatotoxic and non-hepatotoxic molecules within the same class.