Amino azobenzenes are important dyes in the food and textile industry but their application is limited due to their mutagenicity. Computational modeling techniques were used to help understand the factors responsible for mutagenicity, and several quantitative structure toxicity relationship (QSTR) models have been derived. HQSTR (hologram QSTR) analyses indicated that different substituents at sites on both rings contribute to mutagenicity. Fragment parameters such as bond (B) and connectivity(C), as well as donor-acceptor (DA)-based model provide significant results (q(2) = 0.59, r(2) = 0.92, r(2)predictive = 0.63) explaining these harmful effect. HQSTR results indicated that a bulky group at ring "Y" and small group at ring "X" might help to decrease mutagenicity. 3D-QSTR based on comparative molecular field analyses (CoMFA) and comparative molecular similarity index analyses (CoMSIA) are also in agreement with HQSTR. The 3D QSTR studies reveal that steric and electrostatic field effects have a strong relationship with mutagenicity (for CoMFA: q(2) = 0.51, r(2 )= 0.95, r(2) predictive = 0.65 and for CoMSIA: q(2) = 0.51, r(2) = 0.93 and r(2) predictive = 0.84). In summary, negative groups and steric bulk at ring "Y" and small groups at carbon-3 of ring "X" might be helpful in reducing the mutagenicity of azo dyes.