The axiom of extensionality can be weakened, asking only for non-empty sets with the same elements to be equal. Then there can be many "different empty sets" called urelements. I call this theory ZFU (I'm not sure if this coincides exactly with what is denoted by ZFU in the literature).
If we have a model ${\mathfrak A}=(M,\varepsilon)$ of ZFU where $M$ is a set and $\varepsilon$ is a binary relation on $M$, and the set of $\varepsilon$-urelements of $M$ has cardinality $\alpha$, I say that $\mathfrak A$ is an $\alpha$-model of ZFU.
Thus an $1$-model is the same thing as a model of ZFC. Also, if we have an $\alpha$-model of ZFU with $\alpha\neq 0$, then for any urelement $u$, the set of all elements of $M$ whose $\varepsilon$-transitive closure only contains $u$ as an urelement is easily seen to be a model of ZFC.
Thus, the existence of an $\alpha$-model implies the consistency of ZFC. Is the implication strict when $\alpha >1$ ? Also, is the existence of a $\beta$-model a strictly stronger claim than the existence of an $\alpha$-model, when $\alpha<\beta$.
I asked a closely related question here