Extended Thinking
Also: extended thinking mode, Claude thinking mode
Extended thinking was the mode where you explicitly switched Claude into step-by-step reasoning and gave it a fixed token budget (`thinking: {type: "enabled", budget_tokens: N}`). It has been superseded. On Claude Opus 5, Claude Sonnet 5 and Claude Fable 5, thinking is on by default and adaptive — the model decides how deeply to reason — and the extended-thinking configuration is not accepted at all. Depth is now steered with the `effort` parameter. The older configuration still applies on models that support only extended thinking.
In practice
If you have code calling `thinking: {type: "enabled", budget_tokens: 10000}` against `claude-opus-5`, it will fail. Replace it with `output_config: {effort: "xhigh"}` for demanding agentic and coding work, or drop the parameter entirely to get the `high` default. To see the reasoning, which is hidden by default on current models, pass `thinking: {"type": "adaptive", "display": "summarized"}`.
Related concepts
Where Extended Thinking shows up
2 articlesExtended thinking as a toggle is gone on current models — Claude Opus 5, Sonnet 5 and Fable 5 think by default and adaptively, and budget_tokens is not accepted. The control that replaced it is `effort`, which governs total token spend including tool calls. Here's when to raise it, when to lower it, and the two behaviours that will bite you.
You no longer turn thinking on — Claude Opus 5, Sonnet 5 and Fable 5 think by default. The question now is whether you're paying for more deliberation than the task returns. How to set `effort` per workload, and why leaving everything at the default is the most common avoidable cost.