This work revisits the Rhythmicalizer project's approach to classify German free-verse Poetry into prosodic categories using deep learning (Baumann et al., 2018). The rise of Large Language Models (LLMs) and foundation models pre-trained on vast amounts of text and speech begs the question whether these are more appropriate for poetic tasks than the original project's multimodal GRUs. This study evaluates LLM and foundation model performance on the original corpus and compares them to the original classifier.