It didn't use GPT 4 because "This prompt will run on GPT-3.5 turbo instead of GPT-4 as the combination of text and input columns exceed the GPT-4 token limit. To run GPT-4, shorten your prompt or change your model selection in the integration settings."