Tumor vaccines have shown promise in early clinical trials. Among them, tumor cells genetically engineered to secrete biologically active granulocyte-macrophage colony-stimulating factor (GM-CSF) can generate a systemic antitumor immune response. Although the minimal required GM-CSF dose produced by modified tumor cells to achieve a measurable antitumor effect is well known, no data examined whether an upper therapeutic limit may exist for this vaccination strategy. Because recent data demonstrate an immunosuppressive effect of GM-CSF produced by growing tumors, we thus sought to determine whether high GM-CSF doses administered in a vaccine formulation could impair antitumor immunity. Using a vaccine strategy involving a GM-CSF-producing bystander cell line (B78H1-GM) admixed with autologous tumor, we assessed the impact of varying doses of GM-CSF while maintaining a constant antigen dose. Our results defined a threshold above which a GM-CSF-based vaccine not only lost its efficacy, but more importantly for its clinical implications resulted in substantial immunosuppression in vivo. Above this threshold, GM-CSF induced Gr1⁺/CD11b⁺ myeloid suppressor cells that substantially impaired antigen-specific T-cell responses and adversely affected antitumor immune responses in vivo. The dual effects of GM-CSF are mediated by the systemic and not local concentration of this cytokine. Myeloid suppressor cell-induced immunosuppression is mediated by nitric oxide production via inducible nitric oxide synthase (iNOS) because the specific iNOS inhibitor, l-NMMA, restored antigen-specific T-cell responsiveness in vitro. Taken together, our data demonstrated the negative impact of supra-therapeutic vaccine doses of GM-CSF and underscored the importance of identifying these critical variables in an effort to increase the therapeutic efficacy of tumor vaccines.