The development of human cell models stably expressing functional properties of the in vivo cells they are derived from for predicting toxicity of chemicals is a major challenge. For mimicking the liver, a major target of toxic chemicals, primary hepatocytes represent the most pertinent model. Their use is limited by interdonor functional variability and early phenotypic changes although their lifespan can be extended not only by culturing in a 2D dimension under sophisticated conditions but also by the use of synthetic and natural scaffolds as 3D supporting templates that allow cells to have a more stable microenvironment. Hepatocytes derived from stem cells could be the most appropriate alternative but up to now only liver progenitors/hepatoblasts are obtained in vitro. A few hepatocyte cell lines have retained a variable set of liver-specific functions. Among them are the human hepatoma HepaRG cells that express drug metabolism capacity at levels close to those found in primary hepatocytes making them a suitable model for both acute and chronic toxicity studies. New screening strategies are now proposed based on miniaturized and automated systems; they include the use of microfluidic chips and cell chips coupled with high content imaging analysis. Toxicogenomics technologies (particularly toxicotranscriptomics) have emerged as promising in vitro approaches for better identification and discrimination of cellular responses to chemicals. They should allow to discriminate compounds on the basis of the identification of a set of markers and/specific signaling pathways.