Extended & Adaptive Thinking

This skill covers Claude's ability to reason deeply before responding — both the prompt patterns that elicit genuine deliberation (model-agnostic, below) and the current thinking API (how to turn it on and tune its depth). It applies to the current Claude family: Fable 5 (most capable), Opus 4.8 (default), Sonnet 4.6, and Haiku 4.5.

Model IDs and API details move fast. This skill is dated; always confirm against platform.claude.com.

How thinking works today (the API)

The modern control is adaptive thinking plus the effort parameter — not a fixed token budget. The older thinking: {type: "enabled", budget_tokens: N} is deprecated on Opus 4.6 and removed (returns a 400) on Opus 4.7, Opus 4.8, and Fable 5 — do not use it on current models.