The early identification of teratogens in humans and animals is mandatory for drug discovery and development. Zebrafish has emerged as an alternative model to traditional preclinical models for predicting teratogenicity and other potential chemical-induced toxicity hazards. To prove its predictivity, we exposed zebrafish embryos from 0 to 96 h post fertilization to a battery of 31 compounds classified as teratogens or non-teratogens in mammals. The teratogenicity score was based on the measurement of 16 phenotypical parameters, namely heart edema, pigmentation, body length, eye size, yolk size, yolk sac edema, otic vesicle defects, otoliths defects, body axis defects, developmental delay, tail bending, scoliosis, lateral fins absence, hatching ratio, lower jaw malformations and tissue necrosis. Among the 31 compounds, 20 were detected as teratogens and 11 as non-teratogens, resulting in 94.44 % sensitivity, 90.91 % specificity and 87.10 % accuracy compared to rodents. These percentages decreased slightly when referred to humans, with 87.50 % sensitivity, 81.82 % specificity and 74.19 % accuracy, but allowed an increase in the prediction levels reported by rodents for the same compounds. Positive compounds showed a high correlation among teratogenic parameters, pointing out at general developmental delay as major cause to explain the physiological/morphological malformations. A more detailed analysis based on deviations from main trends revealed potential specific modes of action for some compounds such as retinoic acid, DEAB, ochratoxin A, haloperidol, warfarin, valproic acid, acetaminophen, dasatinib, imatinib, dexamethasone, 6-aminonicotinamide and bisphenol A. The high degree of predictivity and the possibility of applying mechanistic approaches makes zebrafish a powerful model for screening teratogenicity.