Extended Thinking Claude Opus: When to Pay for Thinking More

Not all problems require the same level of reasoning. Using the most expensive model for everything is like hiring a senior lawyer to organize your file. Theextended thinkingin Claudeand Opus mode have their place, but knowing when to pay for more thinking—and when not to—makes the difference between investing and wasting.

The case of Andrea: when Opus found what Sonnet did not see

Andrea is a human resources consultant. Complex cases: unfair dismissals, internal conflicts, scattered documentation. A client comes to you with eleven documents, crossed emails that contradict each other, and deadlines.

With normal Sonnet, the response is good, clear and tidy. But too linear for such a tangled case.

With extended thinking, the response improves. Detect nuances. Connect the arguments better.

With Opus what was hidden appears: a contradiction between an email from February and a report from March that completely changes the reading of the dismissal.

Andrea doesn’t think “this AI is smarter.” Think:“This answer is worth money.”

This is noticeable in tasks where there are:

• Conflicts between sources
• Ambiguous criterion
• Decisions with real consequences

On simple tasks, asking you to think more doesn’t add value. It only adds cost.

Thinking more does not mean answering longer. Extended thinking does not consist of longer or more explanatory answers. It consists of the model giving itself more time to reason before responding. The difference is not measured in words. It is measured in connections that did not appear before.

Normal Sonnet vs Extended Thinking Sonnet

Sonnet, as you’ve used it so far, is fast and reliable for most jobs.

When you activate extended thinking, you are telling it: “Before you respond, review your own steps. Don’t go in a straight line.

The improvement is noticeable when:

• The information is contradictory
• The rules are not entirely clear
• Criteria must be prioritized

If the task is mechanical, there is nothing to review. There is no possible improvement there.

The key is to distinguish betweendifficulty and risk. A task may be difficult but low risk. Or simple but with serious consequences if it fails.

Opus: the expensive consultant

Opus is not Sonnet thinking a little more. It’s another profile.

• slower
• Much more expensive
• Able to hold long discussions with many pieces at the same time

Opus shines when the cost of a mediocre response is high. A legal conflict. A strategic decision. An analysis that affects thousands of euros.

Using it to summarize emails or write simple texts is like hiring a senior lawyer to organize your file. It doesn’t contribute anything.

The question that decides the model

Andrea ended up always using the same question before choosing a model:

“If this answer is mediocre, how much does it cost me?”

• If the answer is “little” or “nothing” →normal sonnet
• If the answer is “it can hurt” →Extended thinking
• If the answer is “I can’t afford to fail” →Opus

It is not a technical question. It is an economic question.

A common mistake: upgrading the model due to insecurity

Many users use Opus “just in case.” It’s understandable. It’s also the quickest way to skyrocket the bill without gaining any real quality.

Upgrading only makes sense when there is something to gain. If you don’t know what you are buying with that extra cost, you don’t need it.

The hierarchy is clear:

1. normal sonnetfor 80% of daily tasks: summaries, writing, classification, brainstorming.

2. Sonnet with extended thoughtfor the 15% that requires reviewing steps: analysis with contradictory data, decisions with several variables, tasks where a mistake hurts.

3. Opusfor the 5% where you cannot afford to fail: strategic decisions, legal conflicts, audits with money at stake.

How to know if you are spending well

Choose a real decision that you have pending. Pass it off as normal Sonnet. Repeat it with extended thought. Try Opus. Compare the three answers.

The final question: Does the improvement justify the cost in your case? Write down the answer. That will be your personal rule.

There are professionals who find that 90% of their work works perfectly with normal Sonnet. Others find that extended thinking saves them from costly mistakes several times a week. And some need Opus only once a month, but that time justifies all the cost.

There is no universal answer. There is your answer, based on your decisions and your risks.

Expensive is not better by definition

Don’t make Opus your default model. Expensive is not better by definition. It is better when the problem demands it. Extended thinking and Opus are tools for specific moments, not for every day. Knowing when to use them is as important as knowing how to use them.

Think of it as a team: you have tools for everyday life and tools for critical moments. Using the critical tool for everything does not make you more productive. It makes you poorer without making you wiser.

---