Language-dependent performance of large language models in medical diagnostics: A comparative study of ChatGPT-4o and Claude 3.5 Sonnet | Synapse