A critical obstacle in advanced colorectal cancer (CRC) treatment is the insufficient improvement on survival of conventional chemotherapy. Cancer stem cells are reported to be one of the crucial explanations. CD133 has been identified as a surface marker of CRC stem cells. Bispecific antibodies (BiAbs) targeting tumor-specific antigens are promising therapeutics for malignant diseases, yet that targeting CD133 produced by genetic engineering has not been published. In the current research, CD133 expression in primary CRC was detected by immunohistochemistry, and an asymmetric BiAb consisting of monomer of chimeric AC133 (mouse anti-human CD133 monoclonal antibody) and single chain of humanized OKT3 was developed to eradicate CD133-expressing tumor cells by arming activated T cells in vitro and in vivo. In immunohistochemical examination, CD133 overexpression (>50% of stained cells) frequency was significantly correlated with lymphatic invasion and clinical stage. The new molecular revealed dual-antigen-binding specificity to CD133 and CD3, its distinct structure not only facilitated the purification procedure but also conferred the antibody to ensure a longer and stronger cytotoxic activity. By arming activated T cells, the new antibody displayed impressive cytotoxicity toward CD133(high) but not CD133(low) CRC cells in vitro, produced amounts of cytokines (interferon-γ and granulocyte-macrophage colony-stimulating factor), and could inhibit tumor growth and retard tumor development in nonobese diabetic-severe combined immunodeficient mice without apparent toxicity. Taken together, the new BiAb possesses prosperities that support that the molecule has the potential of being a promising candidate of new therapeutics for CRC therapy.