The amyloid-β protein (Aβ) oligomers are believed to be the main culprits in the cytoxicity of Alzheimer's disease (AD) and p3 peptides (Aβ17-42 fragments) are present in AD amyloid plaques. Many small-molecule or peptide-based inhibitors are known to slow down Aβ aggregation and reduce the toxicity in vitro, but their exact modes of action remain to be determined since there has been no atomic level of Aβ(p3)-drug oligomers. In this study, we have determined the structure of Aβ17-42 trimers both in aqueous solution and in the presence of five small-molecule inhibitors using a multiscale computational study. These inhibitors include 2002-H20, curcumin, EGCG, Nqtrp, and resveratrol. First, we used replica exchange molecular dynamics simulations coupled to the coarse-grained (CG) OPEP force field. These CG simulations reveal that the conformational ensemble of Aβ17-42 trimer can be described by 14 clusters with each peptide essentially adopting turn/random coil configurations, although the most populated cluster is characterized by one peptide with a β-hairpin at Phe19-Leu31. Second, these 14 dominant clusters and the less-frequent fibril-like state with parallel register of the peptides were subjected to atomistic Autodock simulations. Our analysis reveals that the drugs have multiple binding modes with different binding affinities for trimeric Aβ17-42 although they interact preferentially with the CHC region (residues 17-21). The compounds 2002-H20 and Nqtrp are found to be the worst and best binders, respectively, suggesting that the drugs may interfere at different stages of Aβ oligomerization. Finally, explicit solvent molecular dynamics of two predicted Nqtrp-Aβ17-42 conformations describe at atomic level some possible modes of action for Nqtrp.