Extended thinking mode (in chat)

Extended thinking mode is a feature in Claude's chat interface that gives Claude extra time and computational space to reason through problems before delivering a final answer. Instead of switching to a different model, it allows the same model to work through a dedicated internal reasoning process — exploring multiple approaches, checking its own logic, and planning its response — before committing to an output. When extended thinking is enabled, Claude generates 'thinking content blocks' that capture this internal reasoning. Users can see a collapsible 'Thinking' section above Claude's final response, showing a summary of the reasoning process. On Claude 4 models via the API, the full internal reasoning is encrypted for safety, and only a summary is returned. This transparency lets users understand how Claude arrived at an answer and verify the logic. Extended thinking is best suited for tasks where accuracy and depth matter more than speed: complex math, multi-step coding problems, trade-off analysis, research synthesis, and long agentic workflows. It is not universally beneficial — simple or intuitive tasks may not improve and can even suffer from over-reasoning.