The traditional classification and model of acute myeloid leukemia (AML), in common usage for much of the twentieth century, correlates poorly with treatment outcome, biologic studies, and genetic markers in AML, fails to accommodate large subgroups such as typical AML in the elderly or AML following myelodysplastic syndrome (MDS), and (except for acute promyelocytic leukemia) is not used in clinical decisions. Available data suggest an alternative classification and model that initially divides AML into two groups not recognized by traditional classification: MDS-related (MDR)-AML and true de novo (TDN)-AML. MDR-AML includes most AML in the elderly, AML following MDS, AML complicating Fanconi's anemia, and a minor subset of AML in children; these subgroups appear to be linked by a common mutator phenotype, common genetic abnormalities, multilineage hematopoietic dysplasia, clonal hematopoiesis, and poor outcome with cytotoxic chemotherapy. TDN-AML includes AML with the common translocations seen in children and young adults; these subgroups lack features of a mutator phenotype, have approximately flat incidence throughout life, have similar genetic abnormalities, lack multilineage hematopoietic dysplasia and clonal hematopoiesis, and often have good outcome with cytotoxic chemotherapy. Progression in TDN-AML appears to consist predominantly of expansion of a transformed clone, while progression in MDR disease appears to consist initially of progressive accumulation of genetic damage, eventuating in malignant transformation to MDR-AML in some cases. This revised model and classification create therapeutically significant disease groups, allow rapprochement of clinical, morphologic, genetic, and biologic findings in AML, provide a rational model for AML, and frame questions that provide logical direction for future diagnostic, therapeutic, and biologic studies in AML.