Rubric
Contents — domains, guide and mocks

Matching the model to the task

CCAO-F 3.314 min read · checked 21 September 2026

Task statementAlign model selection with task requirements (cost, speed, quality)

How the choice is actually made

  1. Name the constraintwhat would make this unacceptable?
  2. Pick a starting pointcheapest that might work, or most capable
  3. Test on real workyour prompts, your data, your edge cases
  4. Move one stepupgrade if it fails, downgrade if it passes easily

Re-test when the task, the volume or the model lineup changes

The first step is the one that decides everything. If you cannot name the binding constraint, you will default to “the best one”, which is a decision made for you.

Find the binding constraint

Cost, speed and quality trade against each other, but in any real task one of them is usually the thing that would sink it. A customer-facing autocomplete that takes eight seconds is useless however good the text is. A once-a-quarter regulatory analysis that costs three times more is a rounding error. A pilot running on a limited allowance fails if it exhausts the allowance in a week. Naming that constraint out loud converts a vague preference into a decision you can defend.

If the scenario stresses…The binding constraint isWhich points to
“within two seconds”, “while the caller waits”SpeedThe fastest tier, with effort kept low
“40,000 a day”, “every document in the archive”Cost at volumeThe cheapest tier that passes your test
“the regulator will read this”, “contradictory records”Quality on hard reasoningThe capability tier, effort raised
“our allowance ran out on Wednesday”Usage budgetLighter model and lower effort for routine work
“nobody knows if this will work yet”Learning cheaplyPrototype on the efficient tier, then decide

When two constraints genuinely bind — high volume and high stakes together — that is a signal to split the work rather than to compromise on one model. More on that below.

Two documented ways in

Anthropic’s guidance on choosing a model offers two starting strategies rather than a single ranking, and which one you take depends on what you are optimising. Starting efficiency-first means beginning with the fastest, cheapest model — Claude Haiku 4.5 in the current lineup — testing thoroughly and upgrading only if it falls short; it suits prototyping, tight latency requirements, cost-sensitive work and high-volume tasks. Starting capability-first means beginning with the most capable model, optimising your prompts, then downgrading if you can; it suits complex reasoning, accuracy-critical work and advanced technical tasks.

Start cheap and climb, or start strong and trim

Efficiency-first

  • Start with the fastest, cheapest model
  • Test thoroughly against your own cases
  • Upgrade only where it visibly falls short
  • Fits prototypes, volume and tight latency

Capability-first

  • Start with the most capable model
  • Optimise the prompt until quality is right
  • Then downgrade and see what survives
  • Fits complex, accuracy-critical work
Both strategies end in the same place — the lightest model that does the job. They differ in which risk you take first: wasted effort, or wasted money.

The capability-first route has a subtlety worth naming. Optimising the prompt on the strong model and then downgrading often works, because much of what looked like a capability gap was a clarity gap. That is also the honest reason to try it: it separates “this task is genuinely hard” from “my brief was vague”, which is the distinction lesson 1.3 is built on.

Test rather than assume

The documented upgrade-and-downgrade process is not a matter of taste. Create benchmark tests specific to your use case, run them with your actual prompts and data, compare accuracy, quality and handling of edge cases, and weigh the difference against the cost. For a non-technical team this does not require tooling: twenty real examples with known good answers, run on two models, scored on a sheet, will settle an argument that would otherwise run for a month.

Choose the twenty deliberately. Include the awkward ones — the ambiguous invoice, the email in two languages, the case where the right answer is “we cannot tell”. Models diverge most at the edges, and a test set made of easy cases will tell you the tiers are indistinguishable, which is true only of easy cases.

Scoring two tiers on the same twenty cases

  • Passes: Straightforward invoices (12 cases)identical output on both tiers
  • Passes: Multi-currency invoices (3)both correct; cheaper tier slightly faster
  • Check: Missing purchase-order number (2)cheaper tier guesses; stronger tier flags it
  • Fails: Contradictory line totals (2)cheaper tier picks one silently
  • Passes: Scanned, poor quality (1)both decline and ask for a re-scan
A deliberately hard test set. The cheaper tier holds up on everything except the two rows where the reasoning gets ambiguous — which is exactly the information you needed.

Mix models rather than compromise

When volume and difficulty both bind, the documented answer is not to split the difference but to use more than one model. Anthropic describes two patterns. In the executor pattern, a lower-cost model does the work and escalates hard decisions to a frontier model. In the orchestrator pattern, a frontier model plans and delegates the bulk work to lower-cost workers. Both exist to reduce token costs without giving up quality where quality matters.

In an office setting without any code, the same logic appears as a manual routing rule: run the routine pass on the light tier, and send anything it flags as uncertain to a conversation on the capable tier. The decision that makes this work is defining “uncertain” in advance, not after you have seen the output.

The dial inside the model

In the Claude apps there is a second lever that often matters more day to day than the tier. The effort setting controls how thorough a response is and how much of your usage it consumes: low and medium work well for routine tasks and stretch your usage further, high is the default and is described as offering the best overall balance of quality and speed, and extra high and max exist for long-running and deeply demanding work. Thinking is a separate toggle, and on the most capable current models it cannot be switched off.

So the practical version of “align model selection with task requirements” inside the apps is a pair of choices: which model, and how hard it should work. Simple questions and general writing need neither a heavy model nor high effort. Genuinely difficult analysis, planning and technical problems are where raising effort — and, if needed, the tier — is worth the usage it costs. Usage itself is a live constraint: heavy conversations consume allowance faster, Research can consume it faster still because it retrieves many sources, and very long conversations cost more per turn as earlier messages are summarised to keep going.

One last discipline. A model decision is not permanent. Lineups change, prices change, your volumes change, and a choice made against last year’s constraint may now be wrong in both directions. Re-run the twenty cases when something material moves — and keep them, because the test set is the durable asset, not the answer it produced.

Traps the wrong answers are built from

Tempting but wrongDo this instead
Ranking models by quality and always taking the top oneName the binding constraint first, then choose against it.
Deciding between tiers by argumentRun twenty real cases on both and compare on accuracy, edge cases and cost.
Building a test set out of easy casesInclude the ambiguous, contradictory and awkward ones, where tiers actually diverge.
Compromising on one model when volume and stakes both bindRoute routine work to the light tier and escalate defined hard cases.
Changing the model when the allowance is the problemLower effort on routine work, shorten conversations, move standing context into a Project.

You should now be able to

  • Identify the binding constraint — cost, speed or quality — from a described task.
  • Choose between an efficiency-first and a capability-first starting strategy and justify it.
  • Design a small benchmark of real cases, including edge cases, to compare two models.
  • Apply executor and orchestrator style routing when volume and difficulty both bind.
  • Use the effort setting as the within-model dial for thoroughness against usage.
  • Recognise when a usage problem calls for effort and context hygiene rather than a model change.

Practice questions

Original questions written for this lesson, in the exam’s style. Answer first, then open the reasoning — every option is explained, including why the wrong ones are tempting.

  1. Question 1

    A support team wants Claude to suggest a reply while the agent is still on the call. Anything over about three seconds is unusable. The replies are short, drawn from a fixed knowledge base, and the team handles high call volume.

    What is the most appropriate approach?

    1. AUse the most capable model, since reply quality affects customers directly.
    2. BUse the fastest tier with effort kept low, and test it on real call transcripts.
    3. CUse the most capable model with thinking enabled for better judgement.
    4. DAlternate between tiers depending on how busy the queue is.
    Show answer and reasoning
    1. AIncorrect. The heaviest tier is generally slower, and the stated constraint that would sink the feature is latency.
    2. BCorrect. Latency is the binding constraint, the task is straightforward, and the documented efficiency-first route is to start light and verify on real cases.
    3. CIncorrect. Extended reasoning adds latency, which is precisely the constraint the scenario names.
    4. DIncorrect. Variable quality and behaviour during a live call is worse than a consistent choice, and it addresses no stated requirement.
  2. Question 2

    A finance team is arguing about whether the cheaper tier is good enough for month-end variance commentary. One group insists it is fine; the other insists only the top model is safe. Nobody has run the task on both.

    What does the documented guidance suggest they do?

    1. ADefault to the most capable model, since financial output is sensitive.
    2. BSplit the difference by using the middle tier for everything.
    3. CBuild a benchmark from real past months, run both, and compare quality against cost.
    4. DAsk Claude directly which model it would recommend for the task.
    Show answer and reasoning
    1. AIncorrect. It may be the eventual answer, but choosing without evidence is the habit the guidance specifically replaces.
    2. BIncorrect. A compromise chosen to end an argument is not a decision made against the task’s requirements.
    3. CCorrect. Testing with actual prompts and data, comparing accuracy and edge cases, then weighing against cost is the documented process.
    4. DIncorrect. A model’s self-recommendation is not evidence about its performance on this team’s data.
  3. Question 3

    An insurer must process 200,000 claim documents a year. Most are routine, but roughly 4% contain contradictory or ambiguous information where a wrong decision has serious consequences for the customer.

    Which TWO elements should the approach include? (Select 2.)

    1. ARun the routine pass on the efficient tier to keep the volume affordable.
    2. BDefine in advance what counts as ambiguous, and escalate those to the capable tier.
    3. CUse the most capable tier for all 200,000 documents for consistency.
    4. DUse the cheapest tier throughout and accept errors on the 4%.
    5. ELet the model decide case by case which tier to use.
    6. FRaise effort to maximum on every document to be safe.
    Show answer and reasoning
    1. ACorrect. Cost at volume binds on the 96%, and a light tier is documented as suited to high-volume processing.
    2. BCorrect. This is the executor pattern — a lower-cost model escalating hard decisions — and defining the trigger up front is what makes it reliable.
    3. CIncorrect. It multiplies the cost of the routine 96% without improving decisions that were already correct.
    4. DIncorrect. The ambiguous cases are where consequences are serious, which is exactly where capability should be spent.
    5. EIncorrect. The model does not select its own tier, and an undefined escalation rule cannot be audited.
    6. FIncorrect. Maximum effort on 200,000 routine documents consumes budget for no gain on cases that are already clear-cut.
  4. Question 4

    A user on a paid plan keeps hitting usage limits by mid-week. Her work is mostly short rewrites and email drafting, all done in one long-running conversation on the heaviest model at high effort.

    Which change addresses her problem most directly?

    1. AUpgrade to a higher plan so she has more allowance for the same habits.
    2. BMove routine work to a lighter model at lower effort and start fresh conversations.
    3. CTurn thinking on so answers are right first time and need fewer retries.
    4. DUse Research for her drafting so answers are better sourced.
    Show answer and reasoning
    1. AIncorrect. It may help eventually, but it spends money to avoid changing usage that is inefficient for the work described.
    2. BCorrect. Lower effort stretches usage further on routine tasks, and long conversations consume more per turn as earlier messages are summarised.
    3. CIncorrect. Extended reasoning consumes more, not less, and the described tasks are not reasoning-limited.
    4. DIncorrect. Research retrieves many sources and can consume the allowance faster, which worsens the stated problem.

Sources

Drafted with AI assistance and checked against the sources above; expert review is in progress. Spotted an error? Tell us and it gets fixed, dated and listed on how this is written.