Thirteen LLMs, one pipeline, zero excuses. We dig into a controlled study asking whether bigger models actually deliver better ontology learning — or whether the plateau arrives sooner than you think.