Prior studies suggest a possibility that the anticancer property of garlic is more effective only when exposed directly to cancer cells than absorbed first by the normal epithelial cells of the gastrointestinal tract wall. We tested this possibility in two mouse models of highly aggressive malignancies that cannot yet be cured by conventional therapies: sarcoma 180- and EL4-induced lethal ascites. Daily oral gavages of raw garlic extract (RGE; equivalent to 100 mg wet weight) for 21 days failed to offer any meaningful effect in the mice with malignancies. However, the daily injection of the same amounts of the same materials for 21 days completely cured all the mice of cancer. This novel anticancer activity of RGE was present entirely in the size fraction of the molecules smaller than 3000 Dalton rather than the larger molecules and was completely partitioned into the organic phase rather than into the aqueous phase. One half of the anticancer activity was inactivated by heating at 100 °C for 10 min, suggesting that multiple components were concertedly involved. In a direct comparison, the RGE was significantly more effective in killing the cultured cancer cells in vitro than the extracts from other 21 raw vegetables and fruits. In cell culture, RGE killed a wide variety of different cancer cells regardless of species of origin and cell types. Cancer cells generally are well known to be defective in many common metabolic pathways present in their normal cell counterpart for processing normal nutrients. The metabolism of these otherwise normal nutrients could be stalled in the cancer cells and become cytotoxic. The most-effective way of treating cancer by RGE may be the direct injection instead of eating the cooked garlic.