Can I use gmres (or similar) to take direct inverse?

I have a very large (8531x8531), complex, sparse matrix that I need to get the direct inverse of (i.e. not solve system of equations). I need to get the inverse to so some direct manipulation to it and then invert it back. The matrix also has large condition number (condest=6.5E+8). When I take the inverse using inv() or \, I get a residual that is significant enough to impact my analysis (6E-10).

I was thinking of using the gmres function to try to take the inverse directly (i.e. solve AX=B, where X=A^-1 and B=eye). However, this function does not allow b to be a matrix.

I was wondering if anyone could suggest a work around to this problem, or maybe a better way to take the inverse.