Tumor immunotherapy has been an important advancement in cancer treatment in recent years. Compared with T cell-based therapy, natural killer (NK) cell-based therapy does not require human leukocyte antigen matching and has fewer side effects; thus, NK cell therapy has gradually attracted the attention of researchers and clinicians. Reliable and effective animal models are essential for evaluating the effects of NK cell therapy. NK cells kill cancer cells mainly through apoptosis. In this study, we first established a 3D coculture model using fluorescence resonance energy transfer (FRET)-based lung or breast cancer cells and tdTomato-labeled NK cells. We observed that cancer cells changed from green to blue when undergoing apoptosis induced by red NK cells. We then coinjected these green cancer cells with red NK cells into zebrafish to visualize the interaction between them and the killing process of NK cells against cancer cells in real-time and at single-cell resolution in circulation. Using this model, we found that NK cells can quickly kill cancer cells in zebrafish circulation in 40 min and the caspase-3 can be activated in 5–10 min. This FRET-based zebrafish tumor model can serve as a powerful in vivo tool that can facilitate the development of NK cell-based therapy. More importantly, cancer cells from cancer patients can be labeled with our apoptotic biosensor and then transplanted into zebrafish to evaluate the sensitivity of the cancer cells to NK cells to help clinicians make treatment plans that can benefit patients.