fine-tuningllmclassification Fine-Tuning vs Prompting for Intent Classification: A Concrete Comparison We had 8,000 labeled support intents. I ran the same task through GPT-4o prompting and a fine-tuned distilbert. The answer isn't what the blog posts say.