Enhance User Transparency with Summarized AI Reasoning
Access readable 'thinking summaries' to debug or explain AI logic.
Configure the 'thinking.display' attribute to 'summarized' to receive an easy-to-read summary of Claude's reasoning process while using Fable 5 models.
The Scenario
You are building an AI-powered research assistant and want to show users the 'steps' the AI took to reach its conclusion, but the raw output is too long and technical.
Before & after
In previous versions, you either saw the full raw 'Chain of Thought' (which was messy) or nothing at all. Manually parsing or hiding this output for end-users took 10–15 minutes of coding.
Request the 'summarized' thinking mode in the API to get a human-readable digest of the AI's internal reasoning process. This visibility takes about 30 seconds to toggle in your settings.
The Prompt
Update my current Anthropic API request to use Claude Fable 5 with the thinking display set to 'summarized'. Here is my current payload: [PASTE_API_PAYLOAD]
For Claude Fable 5 and Mythos 5, the raw chain of thought is hidden. However, you can set the `thinking.display` parameter to 'summarized'. This provides a clean, readable summary of the model’s internal logic, which is great for debugging or providing transparency to users without overwhelming them with raw data.
Source
Claude Platform - Claude API Docs"set display: "summarized" to receive readable thinking summaries. The raw chain of thought is never returned."