Molecular modeling methods have been used to construct three-dimensional models for agonist and antagonist complexes with beta-adrenergic receptors. The recent rhodopsin crystal structure was used as a template in standard homology modeling methods. The rhodopsin-based homology models were assessed for agreement with experimental results for beta-adrenergic receptors, and compared with receptor models developed using de novo modeling techniques. While the de novo and homology-derived receptor models are generally quite similar, there are some localized structural differences that impact the putative ligand-binding site significantly. The de novo receptor models appear to provide much better agreement with experimental data, particularly for receptor complexes with agonist ligands. The de novo receptor models also yield some interesting and testable hypotheses for the structural basis of beta-adrenergic receptor subtype ligand selectivity.