Because drug-induced liver injury (DILI) remains a major reason for late-stage drug attrition, predictive assays are needed that can be deployed throughout the drug discovery process. Clinical DILI can be predicted with a sensitivity of ~50% and a false positive (FP) rate of ~5% using 24-h cultures of sandwich-cultured primary human hepatocytes and imaging of four cell injury endpoints (Xu et al., 2008). We hypothesized that long-term drug dosing in a functionally stable model of primary hepatocytes (micropatterned cocultures [MPCCs]) could provide for increased predictivity over short-term dosing paradigms. We used MPCCs with either primary human or rat hepatocytes to understand possible species differences along with standard endpoints (glutathione levels, ATP levels, albumin, and urea secretion) to test 45 drugs either known or not known to cause clinical DILI. Human MPCCs correctly detected 23 of 35 compounds known to cause DILI (65.7% sensitivity), with a FP rate of 10% for the 10 negative compounds tested. Rat MPCCs correctly detected 17 of 35 DILI compounds (48.6% sensitivity) and had a higher FP rate than human MPCCs (20 vs. 10%). For an additional 19 drugs with the most DILI concern, human MPCCs displayed a sensitivity of 100% when at least two hepatocyte donors were used for testing. Furthermore, MPCCs were able to detect relative clinical toxicities of structural drug analogs. In conclusion, MPCCs showed superiority over conventional short-term cultures for predictions of clinical DILI, and human MPCCs were more predictive for human liabilities than their rat counterparts.