The notion of 2-topos is the generalization of the notion of topos from category theory to the higher category theory of 2-categories.
There are multiple conceivable such generalizations, depending in particular on whether one tries to generalize the notion of Grothendieck topos or of elementary topos, and in the latter case what axioms one chooses to take as the basis for generalization.
In contrast, (2,1)-toposes are much better understood.
A Grothendieck 2-topos is a 2-category of 2-sheaves over a 2-site.
A Grothendieck (2,1)-topos is a (2,1)-category of (2,1)-sheaves over a (2,1)-site.
See also higher topos theory.
The 2-toposes of 2-sheaves over a 2-site are special among all 2-toposes, in direct generalization of how sheaf toposes (“Grothendieck toposes”) are special among all toposes. In that case, Giraud's theorem famously characterizes sheaf toposes. This characterization has a 2-categorical analog: the 2-Giraud theorem.
A 2-sheaf 2-topos is “$(n,r)$-localic” or “$(n,r)$-truncated” if it has an (n,r)-site of definition.
In particular a $(2,1)$-localic 2-topos is the same as a (2,1)-topos.
Given a 2-topos $\mathcal{X}$, regard it is a 2-site by equipping it with its canonical topology.
Write $Cat(\mathcal{X})$ for the 2-category of internal categories in $\mathcal{X}$, precisely: the 2-category of 2-congruences and internal anafunctors between them (see here).
For $\mathcal{X}$ a 2-topos, there is an equivalence of 2-categories
If $\mathcal{X}$ is $(2,1)$-localic, with a (2,1)-site of definition $C$, then there is already an equivalence
with the 2-category of categories internal to the underlying (2,1)-topos.
If $\mathcal{X}$ is $1$-localic, with 1-site of definition, then there is even already an equivalence
with the internal categories in the underlying sheaf topos.
By the 2-Giraud theorem, $\mathcal{X}$ is an exact 2-category. With this, the first statement is this theorem at 2-congruence.
By the discussion at n-localic 2-topos, a 2-sheaf 2-topos has enough groupoids precisely if it has a (2,1)-site of definition, and has enough discretes precisely if it has a 1-site of definition. With this the second and third statement is this theorem at 2-congruence.
The noteworthy point about theorem 1 is that for an ambient context which is a $(2,1)$-localic (2,1)-topos, the straightforward morphisms of internal categories, hence the notion of internal functors, needs no further localization. This is in stark contrast to the situation for an ambient 1-category. The generalization of this phenomenon is discussed at category object in an (∞,1)-category.
The archetypical 2-topos is Cat. This plays the role for 2-toposes as Set does for 1-toposes.
Given any (2,1)-topos $\mathcal{X}$, the 2-category $Cat(\mathcal{X})$ of internal categories in $\mathcal{X}$ ought to be a 2-topos. But it seems that at the moment there is no proof of this in the literature.
For literature on internal categories in 1-toposes see at 2-sheaf.
2-topos, (∞,2)-topos
An introduction is in
Early developments include
A detailed discussion from the point of view of internal logic is at
Discussion of the 2-categorical Giraud theorem for 2-sheaf 2-toposes is in
Ross Street, Characterization of Bicategories of Stacks Category theory (Gummersbach 1981) LNM 962, 1982, MR0682967 (84d:18006)
Discussion of the elementary topos-analog of 2-toposes is in