Animal models are essential to understanding COVID-19 pathophysiology and for preclinical assessment of drugs and other therapeutic or prophylactic interventions. We explored the small, cheap, and transparent zebrafish larva as a potential host for SARS-CoV-2. Bath exposure, as well as microinjection in the coelom, pericardium, brain ventricle, or bloodstream, resulted in a rapid decrease of SARS-CoV-2 RNA in wild-type larvae. However, when the virus was inoculated in the swim bladder, viral RNA stabilized after 24 h. By immunohistochemistry, epithelial cells containing SARS-CoV-2 nucleoprotein were observed in the swim bladder wall. Our data suggest an abortive infection of the swim bladder. In some animals, several variants of concern were also tested with no evidence of increased infectivity in our model. Low infectivity of SARS-CoV-2 in zebrafish larvae was not due to the host type I interferon response, as comparable viral loads were detected in type I interferon-deficient animals. A mosaic overexpression of human ACE2 was not sufficient to increase SARS-CoV-2 infectivity in zebrafish embryos or in fish cells in vitro . In conclusion, wild-type zebrafish larvae appear mostly non-permissive to SARS-CoV-2, except in the swim bladder, an aerial organ sharing similarities with the mammalian lung.