NK cells are attractive effectors for adoptive immunotherapy of cancer. Results from first-in-human studies with chimeric antigen receptor (CAR)-engineered primary NK cells and NK-92 cells are encouraging in terms of efficacy and safety. To further improve treatment strategies and to test the efficacy of CAR-NK cells in a personalized manner, high-throughput preclinical screening assays using patient-derived tumor samples are needed. Here, we established a flexible Danio rerio (zebrafish) larvae in vivo xenograft model and tested the efficacy of PD-L1-targeting CAR NK-92 cells (PD-L1.CAR NK-92) against the PD-L1-expressing breast cancer cell line MDA-MB-231. We have shown that MDA-MB-231 GFP cells injected into zebrafish larvae at 2 days post fertilization (dpf) are viable and migrate to peripheral parts of the zebrafish body, including the tail region. PD-L1.CAR NK-92 cells injected 2.5 hours later could also migrate to the zebrafish periphery and eliminate cancer cells throughout the body as early as 24 hours, in contrast to parental NK-92 or uninjected controls. Residual cancer cells were further eliminated at later time points, with 48 hours post-injection (hpi) being the best time point for analysis. Confocal live-cell imaging in vivo demonstrated real-time interaction of PD-L1.CAR NK-92 and MDA-MB-231 cells, resulting in cytotoxicity. We further demonstrated that transgenic zebrafish models with labeled blood vessels can be used to study CAR-NK migration through the vasculature. Our data thus suggest that zebrafish larvae can be used for rapid assessment of CAR-NK cell potency in vivo to predict patient response to therapy.