Claude Certified Associate – Foundations free question 1 / 10 · Domain 1: Prompting and Task Execution
A team's problem is slow responses and high cost. Which step, before working on the prompt, does the official documentation say can sometimes be easier?
- Switch to a faster, lower-priced model and compare the results
- Rewrite the evaluation criteria as "fast and cheap" so the criteria count as met
- Increase the number of evaluation items and statistically investigate the cause of the slowness
- Add examples (few-shot) to the prompt to raise output quality
Show the answer and explanation
Answer: A
Why this is correct
The documentation notes that not every success criterion is best solved by prompt engineering, and says that for latency and cost you can sometimes make improvements more easily by choosing a different model. Look at the nature of the problem and decide whether to fix it in the prompt or by changing the model.
Why the other options are wrong
"Rewrite the evaluation criteria as "fast and cheap" so the criteria count as met" … Rewriting the criteria does not make responses faster or cheaper. The documentation says that for some problems, choosing a different model is the easier improvement.
"Add examples (few-shot) to the prompt to raise output quality" … Examples shape the form and quality of the output, but they do not lower response time or cost.
"Increase the number of evaluation items and statistically investigate the cause of the slowness" … More evaluation does not improve speed or cost. The step the documentation names as easier than prompt work is choosing a different model.
Key point
Larger models are generally higher quality but slower and more expensive. The basic split is a lighter model for simple high-volume work and a more capable model for complex reasoning (covered in domain 3).