Based on user feedback on social media platforms and communities, ChatGPT is gradually rolling out a new "Chain-of-Thought (CoT)" layout. By default, it displays the output step by step, but clicking on it will expand the entire chain of reasoning in the right sidebar, with a counter displayed at the top. This change primarily occurs when "Thinking/Reasoning" models or modes are enabled, and is more noticeable on the website. While OpenAI's official release page and recent release notes do not specifically announce the "counter + sidebar" feature, the "Thinking/Reasoning" presentation and UI behavior have been adjusted several times before.
According to public reports, OpenAI is continuing to enhance its presentation of "explainable reasoning" in 2025 (e.g., a more transparent display of o3-mini's thinking process). This interface change aligns with this direction. Currently, this feature is being rolled out in batches, and performance may vary depending on the account and platform. Whether it will become the long-term default style, and the timing of its rollout to mobile and enterprise versions, remains to be determined after official documentation updates and wider deployment.
Frequently Asked Questions
Q: Is this “new layout for thinking output” officially released?
A: It's more like a grayscale launch. Some users have tested it and shared screenshots, but the official documentation hasn't yet specifically explained the specific UI details and counter design.
Q: Where can I see the layout?
A: This mainly occurs in ChatGPT on the web, and is more likely to occur when "thinking/reasoning" models or modes are enabled; different accounts may not be synchronized yet.
Q: What are the uses of the right column and counter?
A: The sidebar displays the complete reasoning chain, and the counter indicates the currently expanded/total steps, which is convenient for backtracking and comparison.
Q: Does this mean that the original CoT will be publicly displayed?
A: It's not the same as "original thought leakage." Media reports have previously stated that OpenAI will be more transparent about its reasoning, but it may still be filtered or summarized.
Q: Does the official have any relevant background information?
A: Recent release notes and reports indicate that the inference experience is continuously being upgraded, but there is no dedicated entry for the details of this "item-by-item display + counter".
