In the present work, a set of ligand- and receptor-based 3D-QSAR models were developed to explore the structure–activity relationship of 109 benzimidazole-based interleukin-2-inducible T-cell kinase (ITK) inhibitors. In order to reveal the requisite 3D structural features impacting the biological activities, a variety of in silico modeling approaches including the comparative molecular field analysis (CoMFA), comparative molecular similarity indices analysis (CoMSIA), docking, and molecular dynamics were applied. The results showed that the ligand-based CoMFA model (Q2 = 0.552, R2ncv = 0.908, R2pred = 0.787, SEE = 0.252, SEP = 0.558) and CoMSIA model (Q2 = 0.579, R2ncv = 0.914, R2pred = 0.893, SEE = 0.240, SEP = 0.538) were superior to other models with greater predictive power. In addition, a combined analysis between the 3D contour maps and docking results showed that: (1) Compounds with bulky or hydrophobic substituents near ring D and electropositive or hydrogen acceptor groups around rings C and D could increase the activity. (2) The key amino acids impacting the receptor–ligand interactions in the binding pocket are Met438, Asp500, Lys391, and Glu439. The results obtained from this work may provide helpful guidelines in design of novel benzimidazole analogs as inhibitors of ITK. © 2013 Wiley Periodicals, Inc.