IL-2 has been used for the treatment of melanoma and renal cell carcinoma, but this therapy has limited efficacy and severe toxicity. Currently, it is assumed that part of the limited efficacy is due to the IL-2-driven preferential expansion of regulatory T cells, which dampen the antitumor immunity. In this study, we characterize a human IL-2 mutant with higher antitumor efficacy and lower toxicity than wild type human IL-2 (wtIL-2). The mutant differs from wtIL-2 by four mutations at the interface with the α subunit of IL-2R. The IL-2 mutant induces in vitro proliferation of CD8(+)CD44(hi) and NK1.1 cells as efficiently as does wtIL-2, but it shows a reduced capacity to induce proliferation of CD4(+)Foxp3(+) regulatory T cells. The IL-2 mutant shows a higher antimetastatic effect than does wtIL-2 in several transplantable tumor models: the experimental metastasis model of MB16F0 melanoma and the experimental and spontaneous metastasis models for the mouse pulmonary carcinoma 3LL-D1222. Relevantly, the IL-2 mutant also exhibits lower lung and liver toxicity than does wtIL-2 when used at high doses in mice. In silico simulations, using a calibrated mathematical model, predict that the properties of IL-2 mutein are a consequence of the reduction, of at least two orders of magnitude, in its affinity for the α subunit of IL-2R (CD25). The human IL-2 mutant described in the present work could be a good candidate for improving cancer therapy based on IL-2.