Recurrent mutations at R140 and R172 in isocitrate dehydrogenase 2 (IDH2) occur in many cancers, including ∼12% of acute myeloid leukemia (AML). In preclinical models these mutations cause accumulation of the oncogenic metabolite R-2-hydroxyglutarate (2-HG) and induce hematopoietic differentiation block. Single-agent enasidenib (AG-221/CC-90007), a selective mutant IDH2 (mIDH2) inhibitor, produced an overall response rate of 40.3% in relapsed/refractory AML (rrAML) patients with mIDH2 in a phase 1 trial. However, its mechanism of action and biomarkers associated with response remain unclear. Here, we measured 2-HG, mIDH2 allele burden, and co-occurring somatic mutations in sequential patient samples from the clinical trial and correlated these with clinical response. Furthermore, we used flow cytometry to assess inhibition of mIDH2 on hematopoietic differentiation. We observed potent 2-HG suppression in both R140 and R172 mIDH2 AML subtypes, with different kinetics, which preceded clinical response. Suppression of 2-HG alone did not predict response, because most nonresponding patients also exhibited 2-HG suppression. Complete remission (CR) with persistence of mIDH2 and normalization of hematopoietic stem and progenitor compartments with emergence of functional mIDH2 neutrophils were observed. In a subset of CR patients, mIDH2 allele burden was reduced and remained undetectable with response. Co-occurring mutations in NRAS and other MAPK pathway effectors were enriched in nonresponding patients, consistent with RAS signaling contributing to primary therapeutic resistance. Together, these data support differentiation as the main mechanism of enasidenib efficacy in relapsed/refractory AML patients and provide insight into resistance mechanisms to inform future mechanism-based combination treatment studies.