Evaluating the Progression of Large Language Model Capabilities for Small-Molecule Drug Design — ThinkLLM