CCAR-P sample question: choosing a model under latency and budget limits

An original sample question by CERTain·exam, based only on Anthropic's public documentation. CERTain·exam is not affiliated with Anthropic.

Domain 2 · Claude Models, Prompting & Context Engineering (13% of CCAR-P in the exam guide effective until December 6, 2026; 13% from December 7, 2026). Objective covered: "Select appropriate Claude models based on trade-offs."

This question tests whether you match the starting model to the constraints in the scenario: latency, volume, a fixed unit budget and the kind of task. Work through it before you reveal the answer.

The question

Quillon Travel is adding suggested replies for human chat agents. A suggestion must appear within about one second while the agent is still reading the customer's message. Volume is about four million suggestions a month, and finance has fixed the budget per suggestion from day one. Each suggestion is a short, routine reply of one or two sentences based on a policy snippet supplied in the request. The team has an evaluation set of 500 past chats with approved replies.

Which starting approach should the architect recommend? (Select one.)

  • A. Start with the fastest, lowest-cost tier, test it on the evaluation set, and upgrade only where it falls short.
  • B. Start with a top-capability model, test it on the evaluation set, and step down a tier later once costs are reviewed.
  • C. Use an Opus-class model in fast mode so replies stream faster, accepting a higher price per request.
  • D. Start with the balanced mid-tier model for all suggestions, because routine replies still benefit from reasoning headroom, and judge quality from agents' feedback after launch.
Show the answer and explanation

Answer

A. Start with the fastest, lowest-cost tier, test it on the evaluation set, and upgrade only where it falls short.

How to read this question

  1. Situation. About one second to show a reply, four million requests a month, a budget per suggestion fixed from day one, short routine replies grounded in a supplied policy snippet, and an evaluation set of 500 approved chats.
  2. What is asked. The starting approach, not the final architecture. One answer.
  3. Option by option.
    • A (correct). Tight latency, high volume, a fixed unit budget, and a straightforward, grounded task are exactly the cases the documentation lists for starting efficiency-first. The evaluation set confirms quality before any upgrade.
    • B. A capability-first start (the page begins with Opus 5.5 and moves to Fable 5.1 only if evals fall short) is a valid strategy for complex reasoning or when accuracy outweighs cost. Here, the replies are routine and the budget is fixed from day one, so starting at the top tier conflicts with both cost and latency. "Most workloads start with Claude Opus 5.5" describes the general default; the same page lists tight latency, cost sensitivity, and high-volume straightforward tasks as the cases for efficiency-first, and all three apply here.
    • C. C fails on cost first. Fast mode (available only on Opus-class models) costs more per request, and the fixed unit budget rules that out. It also speeds up output tokens, not time to first token, and output speed matters little for a one- or two-sentence reply.
    • D. The mid tier is not justified by any fact in the scenario, and judging quality from feedback after launch skips the evaluation set the team already has.
  4. Rule to remember. When latency, volume and a fixed unit cost all point the same way and the task is simple and grounded, start with the fastest, lowest-cost model and let your evaluation set tell you where to upgrade.

Sources (Claude Docs, quotes checked Oct 23, 2026; page checked Oct 27, 2026)

  • Choosing a model:
    • "For many applications, starting with a faster, more cost-effective model like Claude Haiku 5.5 can be the optimal approach:" with steps "Test your use case thoroughly." / "Upgrade only if necessary for specific capability gaps."
    • Efficiency-first is best for "Applications with tight latency requirements", "Cost-sensitive implementations", "High-volume, straightforward tasks".
    • Model matrix: "The lowest latency and price" → "Real-time applications, high-volume intelligent processing, cost-sensitive deployments".
    • Capability-first is best for "Complex reasoning tasks" and "Applications where accuracy outweighs cost considerations".
    • "Most workloads start with Claude Opus 5.5."
    • "Create benchmark tests specific to your use case - having a good evaluation set is the most important step in the process." and "Test with your actual prompts and data."
  • Fast mode:
    • "Fast mode delivers up to 2.5x higher output tokens per second from Claude Opus 5.5, Claude Opus 5, and Claude Opus 4.8 at premium pricing."
    • "Speed benefits are focused on output tokens per second (OTPS), not time to first token (TTFT)".

More practice

Questions

Where does this question come from?

It's an original sample question that CERTain·exam wrote based on Anthropic's public documentation and the public exam guide. Exam content is confidential under Anthropic's candidate agreement, and we do not use it.

What does CCAR-P Domain 2 cover?

According to the exam guide effective until December 6, 2026, Claude Models, Prompting & Context Engineering (13%) covers choosing Claude models based on trade-offs, system prompts, templates and guardrails, prompt engineering techniques, context window and token management, and prompt reuse (caching, modular prompts, Skills). From December 7, 2026, the domain keeps its 13% weight, and its objectives are rewritten. See what changes on December 7, 2026.

Get your free practice exam

First, sign in with your email.