Accurate predication of hepatotoxicity is a major problem in pharmacology that is only partially addressed by current testing protocols. Animal hepatotoxicity testing is expensive, unsuited to high throughput and overall has unreliable concordance with human hepatotoxicity, while standard in vitro systems have, at best, marginally improved productivity. We hypothesize that the adaptation of the sequentially layered, self-assembly liver model (SQL-SAL), developed by the University of Pittsburgh Drug Discovery Institute, into MIMETAS’ high-throughput microfluidic OrganoPlate®, will improve the predictivity over conventional 2D and animal models. The resulting platform contains 96 x 3D microfluidic co-cultures of human primary or iPS hepatocytes with three non-parenchymal human liver cell types. In OrganoPlates™, extracellular matrix (ECM) gels can be freely patterned in microchambers through the use of the PhaseGuide™ technology. PhaseGuides™ (capillary pressure barriers) define channels within microchambers that can be used for extracellular matrix deposition or medium perfusion. The microfluidic channel dimensions not only allow solid tissue and barrier formation, but also perfused tubular endothelial vessel structures (vascularization) can be grown in the medium perfusion channel. Utilizing the PhaseGuide™ technology we are able to engineer our culture to mimic the structure of the liver sinusoid by culturing hepatocyte and stellate cells in an extracellular matrix protein gel, fed by microfluidic nutrient perfusion from an adjacent endothelial and Kupffer cell-lined blood vessel mimic. We report long-term maintenance of metabolic activity by measurement CYP3A4, and liver specific function by albumin production. Per million cells, the OrganoPlate model produces greater than 800ng albumin/day. We also report multi-parameter high-content imaging based readouts of toxicity including mitochondrial function and steatosis. Testing of the known hepatotoxin acetaminophen shows a more physiologically-relevant sensitivity to the drug in the OrganoPlate culture system, with an IC-50 in the sub-millimolar range, much closer to the in vivo human Cmax than 2D culture, where hepatotoxicity is not observed until the millimolar range. These studies display the feasibility of using our 3D human model as a high-throughput screening platform for assessment of pharmaceutical and environmental hepatotoxicity.