Fifty derivatives of phenol whose toxicity is reported have been studied for Quantitative Structure Toxicity Relationship (QSTR). With the help of seven descriptors viz. Core-core repulsion, electronic energy, Log P, dielectric energy, softness, molar refractivity and steric energy. The values of these descriptors have been evaluated with the help of PC MODEL software using the semi empirical PM3 Hamiltonian. 81 multi linear regression models using different combination of descriptors have been developed using the project leader program of CAChe software. The predicted toxicity is close to the observed toxicity and the correlation coefficient in eight models is above 0.80 and, in twenty-nine models above 0.70 and in the remaining thirty have values below 0.5. The values are between 0.50-0.69. 14 models the best QSTR models is obtained by using combinations of descriptors, core-core repulsion, electronic energy, Log P and dielectric energy and the second best by using the combination of descriptors, core-core repulsion, electronic energy, softness and Log P. These models can be treated as very reliable models for predicting the toxicity of phenol derivatives and in helping the preparation of compound of desired toxicity.