Accurately predicting the structure of a protein has been a long standing and extremely challenging problem in biology. In recent years, the rapid evolution and adoption of artificial intelligence (AI) in scientific domains including biology have made the prediction of protein structures leveraging deep learning (DL) frameworks with accuracy rivaling that of experimental crystal structures possible. These advances are key to understanding protein function and play a central role in accelerating the drug discovery process. This work focuses on comparing the performance of state-of-the-art protein structure prediction models across a predefined set of 7 challenging biomolecule categories evaluated using AlphaFold2, AF2Complex, AlphaFold-multimer, AlphaFold3, Chai-1, and Boltz-1. The results compare the accuracy and performance of each method when applied to classes of challenging protein-protein complexes. The evaluation was conducted leveraging computational resources at Oak Ridge National Laboratory (ORNL) including Frontier, ORNL’s exascale supercomputer. The data set constructed, the resulting biomolecular complex type classification, and the comprehensive set of guidelines derived can aid in future experiment design. The results from this study can also provide a deeper understanding of the advantages and limitations of each model when applied to specific classes of biomolecular complexes as opposed to individual complexes.