Question 16 — AB-620 Practice Exam 1
A team is deciding between several Prompt builder models for a complex classification task. The required classification quality cannot be determined reliably from the model category alone. What should the team do before choosing the production model? (Select one answer)
Choose the model with the highest billing rate because it guarantees the best result
Use the default model because structured tasks must remain on GPT-4.1 mini
Test the prompt only after it has been added to the production topic
Run the prompt with representative inputs against the candidate models and compare the results
Correct answers
Run the prompt with representative inputs against the candidate models and compare the results
Explanations
Correct:
Run the prompt with representative inputs against the candidate models and compare the results
Testing the same prompt with candidate models is the reliable way to determine which model delivers the required quality for the task. Model category provides guidance about capability and cost, but actual task performance should be validated.
Wrong:
Choose the model with the highest billing rate because it guarantees the best result
A higher-cost model may provide stronger reasoning capability, but cost alone does not establish that it will produce the required output quality for a specific prompt.
Use the default model because structured tasks must remain on GPT-4.1 mini
GPT-4.1 mini is the default, but Prompt builder allows other managed models when the task requires different capabilities.
Test the prompt only after it has been added to the production topic
Prompt builder provides a testing environment before the prompt is added to a topic. Testing earlier makes prompt and model problems easier to isolate and correct.