In this study the efficiency of each type of CODESSA descriptors to correlate with benzodiazepines toxicity was investigated. The AM1 semiempirical approach was used to computationally optimize 26 different benzodiazepine drugs. Subsequent QSAR investigations at HF/3-21G level using CODESSA package was employed to correlate the drugs’ potency with quantum chemical, electrostatic, geometrical, topological and constitutional descriptors related to each of the structures. The concluded QSAR models reflected that the drugs toxicity was mainly attributed to both geometrical and topological descriptors without any electrostatic contribution. Statistically, the most significant overall correlation was a five-parameter equation with good statistical parameters; correlation coefficient, R2= 0.959, cross-validated correlation coefficient, R2 CV = 0.901, Fisher F-criterion, F = 93.87, and the standard deviation of the regression, S2 = 0.0242. The obtained model is good enough to be used to estimate the activities of the benzodiazepines drugs toxicity.