I would use cond. This gives you a numerical estimate of how close to singular a matrix is (where Inf is a singular matrix).
For example:
m = randn(4);
cond(m) %Well conditioned, usually in the 10's
m = diag([1e-6 1 2 1e6]);
cond(m) %Less well conditioned, 1e12
m = diag([0 1 2 3]);
cond(m) %Singular: Inf