Proarrhythmia models use electrophysiological markers to predict the risk of torsade de pointes (TdP) in patients. The set of variables used by each model to predict the torsadogenic propensity of a drug has been reported to correlate with clinical outcome; however, these reports should be interpreted cautiously as no model has been independently assessed. Each model is discussed along with its merits and shortcomings; none, as yet, having shown a predictive value that makes it clearly superior to the others. As predictive as these models may become, extrapolation of results directly to the clinic must be exercised with caution. The use of in silico models, from subcellular to whole system, is rapidly beginning to form the first line of screening activity in many drug discovery programmes, indicating that biological experimentation may become secondary to analysis by simulation. In vitro proarrhythmia models challenge current perceptions of appropriate surrogates for TdP in man and question existing non-clinical strategies for assessing proarrhythmic risk. The rapid emergence of such models, compounded by the lack of a clear understanding of the key proarrhythmic mechanisms has resulted in a regulatory reluctance to embrace such models. The wider acceptance of proarrhythmia models is likely to occur when there is a clear understanding and agreement on the key proarrhythmia mechanisms. With greater acceptance and ongoing improvements, these models have the potential to unravel the complex mechanisms underlying TdP.