Originally published at claudeguide.io/claude-extended-thinking-guide

Claude Extended Thinking: How to Use Thinking Mode (2026 Guide)

Claude Extended Thinking gives the model a private scratchpad to reason through a problem before returning its final answer. You set a budget_tokens parameter — typically 10,000 to 32,000 tokens — that controls how long it can think. Anthropic benchmarks show ~30% accuracy improvement on hard math and multi-step logic tasks versus standard mode. The model emits one or more "thinking blocks" (internal chain-of-thought) before the visible response. Extended Thinking is available on claude-sonnet-4-5 and claude-opus-4 models.

---

How Extended Thinking works

The mental model is straightforward: instead of asking Claude to produce an answer in a single forward pass, you give it permission to "think out loud" in a hidden buffer. That buffer is bounded by budget_tokens.

The two response block types

When thinking is enabled, the API response contains an array of content blocks that can include two types:

Block type	What it contains	Visible to end users?
thinking	Internal reasoning — hypotheses, scratch calculations, dead ends	No (you choose)
text	The final, polished answer	Yes

The thinking block text is the model's unfiltered reasoning chain. It may include self-corrections like "wait, that contradicts what I said earlier — I need to redo this." That internal revision is precisely why accuracy improves: the model catches its own mistakes before committing to an answer.

The budget_tokens parameter

budget_tokens is the upper limit on how many tokens Claude can spend on internal reasoning. It is not a guarantee — the model may use fewer tokens if the problem is simpler than expected.

Practical ranges as of May 2026:

Budget	Best for
1,000–2,000 tokens	Straightforward reasoning, moderate code tasks
5,000–10,000 tokens	Complex math proofs, multi-step legal analysis
16,000–32,000 tokens	Hard competition-style problems, large-codebase debugging

For a real-world example of how model selection and feature gating cut API bills by 85%, see the Claude API cost case study.

---

Comparison: with vs. without thinking on the same task

To make the tradeoff concrete, here is a side-by-side on a multi-step logic problem.

Prompt: "Three friends — Alice, Bob, and Carol — each have a different number of siblings: 1, 2, and 3. Alice does not have the most siblings. Bob has more siblings than Carol. Who has 3 siblings?"

Without thinking (standard mode):

Bob has 3 siblings.

Correct — but no audit trail. If the problem were harder, you would not know where the reasoning failed.

With thinking enabled (budget_tokens=1000):

Thinking block:

Let me assign variables. The numbers are 1, 2, 3.
Alice does not have the most → Alice ≠ 3.
Bob 

If you are looking to improve the prompts that feed into thinking mode, the **[P1 Power Prompts 300](https://shoutfirst.gumroad.com/l/agfda)** library includes 300 tested prompt patterns optimized for Claude's extended reasoning — covering math, code, legal, and research domains.