Bacillus thuringiensis (Bt) strains produce pore-forming toxins (PFTs) that attack insect pests. Information for pre-pore and pore structures of some of these Bt toxins is available. However, for the three-domain (I-III) crystal (Cry) toxins, the most used Bt toxins in pest control, this crucial information is still missing. In these Cry toxins, biochemical data have shown that 7-helix domain I is involved in insertion in membranes, oligomerization and formation of a channel lined mainly by helix α4, whereas helices α1 to α3 seem to have a dynamic role during insertion. In the case of Cry1Aa, toxic against Manduca sexta larvae, a tetrameric oligomer seems to precede membrane insertion. Given the experimental difficulty in the elucidation of the membrane insertion steps, we used Alphafold-2 (AF2) to shed light on possible oligomeric structural intermediates in the membrane insertion of this toxin. AF2 very accurately (<1 Å RMSD) predicted the crystal monomeric and trimeric structures of Cry1Aa and Cry4Ba. The prediction of a tetramer of Cry1Aa, but not Cry4Ba, produced an ‘extended model’ where domain I helices α3 and α2b form a continuous helix and where hydrophobic helices α1 and α2 cluster at the tip of the bundle. We hypothesize that this represents an intermediate that binds the membrane and precedes α4/α5 hairpin insertion, together with helices α6 and α7. Another Cry1Aa tetrameric model was predicted after deleting helices α1 to α3, where domain I produced a central cavity consistent with an ion channel, lined by polar and charged residues in helix α4. We propose that this second model corresponds to the ‘membrane-inserted’ structure. AF2 also predicted larger α4/α5 hairpin n-mers (14 ≤n ≤ 17) with high confidence, which formed even larger (~5 nm) pores. The plausibility of these models is discussed in the context of available experimental data and current paradigms.