However, the reality is more complex. A 2024 study by Mount Sinai tested several LLMs, including GPT-4, on coding tasks. Even the best-performing model GPT-4 achieved just 45.9% accuracy for ICD-9 ...