Extended thinking

Extended thinking is Anthropic's flag on Claude that allocates a configurable budget of internal reasoning tokens before the user-visible answer — enabling deeper reasoning on hard problems for a higher cost.

Anthropic's equivalent of OpenAI's o-series reasoning. When you enable extended thinking with a token budget (e.g. 8,000 tokens), Claude runs an internal chain-of-thought that the user never sees before emitting the final answer. The technique materially improves performance on math, code, planning, and multi-step reasoning. By 2026 it's a per-call setting on Claude 4.x. Best practice: route by task — disable extended thinking for chat and extraction, enable with a 4K-8K budget for hard reasoning, escalate to 16K only when measured quality warrants the cost.

When to use extended thinking

• Math, code, planning, multi-step reasoning.
• Agent step decisions where one wrong step cascades.

Common mistakes

• Adding explicit "think step by step" instructions when extended thinking is enabled — often hurts quality.
• Running every query with extended thinking — costs and latency blow up.

FAQ

Related terms

• Reasoning model — A reasoning model is an LLM trained to produce extensive internal chain-of-thought before its final answer, trading latency for higher accuracy on hard problems.
• Reasoning tokens — Reasoning tokens (or thinking tokens) are the internal chain-of-thought tokens reasoning models produce before the user-visible answer — billed separately and not shown to the end user.
• Chain-of-thought prompting — Chain-of-thought (CoT) prompting tells a language model to write its reasoning steps before its final answer, increasing accuracy on multi-step problems.

Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/extended-thinking.md.