Thinking Mode

Thinking mode is a provider-specific setting that lets a model spend dedicated computation on intermediate reasoning before producing its answer. It is commonly exposed as thinking blocks, hidden reasoning, a token budget, or an effort control.

Thinking mode is not the same as asking the model to show chain-of-thought reasoning. Modern systems may use hidden or summarized reasoning while returning only a concise final answer. The important operational effect is that the model can allocate more work to decomposition, verification, and tool-use decisions.

Thinking mode affects:

  • output-token usage and billing;
  • latency;
  • context window pressure;
  • prompt caching behavior;
  • tool-use loops; and
  • whether prior reasoning state must be preserved across turns.

It should be used deliberately. For simple tasks, thinking mode can waste tokens. For complex coding, math, analysis, or long-horizon tasks, it can improve reliability if the application measures outcomes.

Anthropic documents thinking configuration, token accounting, and tool-use behavior in its Thinking guide.

The LLM Knowledge Base is a collection of bite-sized explanations for commonly used terms and abbreviations related to Large Language Models and Generative AI.

It's an educational resource that helps you stay up-to-date with the latest developments in AI research and its applications.

© 2026 Promptmetheus