Systematic studies of the intrinsic irreversible strain limit ε irr,0 , microstructure, and microchemistry were made on several internal-tin Nb 3 Sn pre-production wires, fabricated for the domestic agencies of the USA and China participating in the International Thermonuclear Experimental Reactor. These wires were produced by Luvata, Oxford Superconducting Technology (OST), and Western Superconducting Technologies (WST), and were intended for the tokamak's toroidal-field coils. The results of this study show that, for a final heat-treatment at 650 °C to form the A15 phase, both ε irr,0 and the de-pinning field B c2 * improved by increasing heat-treatment duration beyond 100 h for the Luvata wires. On the other hand, we saw no improvement in these two parameters as a function of heat-treatment duration in the OST wires. Furthermore, micro-chemical analysis of OST wires revealed that some Nb 3 Sn filaments have a Sn-and Ti-rich phase at the interface between Cu(Sn) matrix and Nb 3 Sn in the form of a shell around individual filaments. This phase is far less prominent in the Luvata and WST conductors, and could inhibit diffusion of Sn and Ti into Nb 3 Sn filaments during the reaction and may potentially be the reason for the lack of noticeable change in B c2 * with heat-treatment duration in the OST wires. The increase of ε irr,0 and B c2 * with heat-treatment duration in the Luvata wires and the lack of increase in the OST wires may suggest a possible correlation between ε irr,0 and the stoichiometry of the A15 composition. Investigation of the samples' microstructure revealed only a small number of cracked Nb 3 Sn filaments despite the significant and permanent degradation of their critical current I c when subjected to longitudinal tensile strain ε beyond ε irr,0 . The scarcity of cracks indicate that I c (ε) measurements are highly sensitive to crack formation in Nb 3 Sn filaments, especially at low electric-field criteria ⩽0.1 μV cm −1 , even when the sizes of the individual filaments are only few micrometers. All the strands contained substantial Kirkendall porosity, but we found that the quantity and distribution of the Kirkendall voids vary significantly with strand design. Luvata wires have the least porosity, followed by WST wires,