The OECD test guideline 203 for determination of fish acute toxicity requires substantial numbers of fish and uses death as an apical end point. One potential alternative are fish cell lines; however, several studies indicated that these appear up to several orders of magnitude less sensitive than fish. We developed a fish gill cell line-based (RTgill-W1) assay, using several measures to improve sensitivity. The optimized assay was applied to determine the toxicity of 35 organic chemicals, having a wide range of toxicity to fish, mode of action and physicochemical properties. We found a very good agreement between in vivo and in vitro effective concentrations. For up to 73% of the tested compounds, the difference between the two approaches was less than 5-fold, covering baseline toxicants but as well compounds with presumed specific modes of action, including reactivity, inhibition of acetylcholine esterase or uncoupling of oxidative phosphorylation. Accounting for measured chemical concentrations eliminated two outliers, the hydrophobic 4-decylaniline and the volatile 2,3-dimethyl-1,3-butadiene, with an outlier being operationally defined as a substance showing a more than 10-fold difference between in vivo/in vitro effect concentrations. Few outliers remained. The most striking were allyl alcohol (2700-fold), which likely needs to be metabolically activated, and permethrin (190-fold) and lindane (63-fold), compounds acting, respectively, on sodium and chloride channels in the brain of fish. We discuss further developments of this assay and suggest its use beyond predicting acute toxicity to fish, for example, as part of adverse outcome pathways to replace, reduce, or refine chronic fish tests.