XNews

OpenAI ChatGPT Feature Updates

OpenAI ChatGPT announced a set of targeted feature updates focused on user control over GPT-5 Thinking pacing, improvements to in-chat search, and highlights related to the role of memory in conversational quality. The most concrete change is the ability for ChatGPT Plus, Pro, and Business subscribers to select Thinking pace for GPT-5 with Thinking on the web: Standard (new default, balancing speed and intelligence) and Extended (previous default for Plus). Pro users get two additional options, Light (snappiest) and Heavy (deeper). The pace selection persists across web sessions until changed. Separately, OpenAI rolled out search improvements intended to reduce hallucinations, better detect shopping intent (showing product results when appropriate and avoiding them when not), and provide clearer, more usable answer formatting. User commentary from independent posts underscores the practical impact of related settings: one user reported that enabling Developer Mode (which disables memory by default) made ChatGPT feel less capable, reinforcing that memory is a differentiator in perceived intelligence. Together, these announcements reflect a dual emphasis by OpenAI on granularity of user control (letting users trade speed for depth) and on improving factual reliability and task relevance in search. The updates are immediately relevant to paying tiers (Plus/Pro/Business) and have implications for user experience, developer debugging, and trust in conversational accuracy. Sources include OpenAI’s official announcement and user posts reporting observed changes and behavior related to memory and search, providing both product-side detail and real-world user perspective.

Key insights

Product Team Perspective (OpenAI Announcement)

OpenAI frames the update as delivering user agency and improved reliability. By exposing GPT-5 Thinking pace controls to Plus, Pro, and Business users on web, the company enables users to select trade-offs between speed and depth of reasoning. The new Default (Standard) aims to balance latency and intelligence while Pro-only Light and Heavy options allow finer granularity. Parallel search improvements address hallucination reduction, shopping intent detection, and answer formatting—three concrete areas that can be measured and iterated on. These changes indicate a product roadmap focused on configurable performance, tiered feature differentiation, and incremental reliability gains rather than wholesale model changes.

User Feedback Perspective

User reports corroborate the product changes and reveal downstream perception effects. One user documented that Developer Mode disabled memory and led to a noticeable drop in conversational fluency, suggesting that memory is central to perceived intelligence. Other commentary emphasizes the practical benefits of search improvements—fewer hallucinations and better shopping detection reduce friction for commerce-related queries. Collectively, users value both responsiveness and contextual continuity; allowing choice over thinking pace empowers users to prioritize the aspect most relevant for a session, while search upgrades address day-to-day reliability.

Industry/Analyst Perspective

Analysts would interpret these updates as iterative but strategic: pace controls are a low-friction way to let customers tailor AI behavior without retraining, while search fixes tackle concrete failure modes (hallucinations, intent recognition) that impact enterprise adoption. The persistence of settings on web signals attention to user preferences and session continuity. The memory-related user observation highlights that stateful features remain an important moat for chat products. Overall, this signals OpenAI’s focus on product polish, tiered differentiation, and incremental trust-building—factors important for retention in competitive conversational AI markets.

Source posts