There's the common random string model (where hash functions can be modeled as having been picked from a family of functions using public coins).
There are also "whatever-tractable random oracles", where adversaries also have an oracle that finds a whatever with respect to the random oracle.
(Usually 'whatever' is one of {'preimage','second-preimage','collision'}.)
There's also something I've thought of that I've never heard of actually being used, where the oracle is drawn from a distribution of oracles such that, for any algorithm that adaptively makes a feasible number of queries to the oracle and then outputs a guessed $\:\langle \hspace{.01 in}x\hspace{.01 in},\hspace{-0.02 in}y\rangle\:$ pair, the probability that
the algorithm did not query the oracle on $\hspace{.01 in}x$ $\:$ and $\:$ the oracle's output on $x$ is $y$
is negligible. (Perhaps the "Unpredictable Oracle Model"?)