OpenAI has rolled out Dreaming V3, a rebuilt memory system for ChatGPT that automatically learns and recalls details about users with 82.8% accuracy, while consuming roughly five times less computing power than the previous approach. The update, which began rolling out to ChatGPT Plus and Pro subscribers in the United States on June 4, 2026, represents a fundamental shift in how AI assistants remember users: instead of requiring people to manually save facts, the system now synthesizes information from past conversations on its own.

What Changed in ChatGPT's Memory System?

The original ChatGPT memory feature, launched in February 2024, worked like a digital notepad. Users had to explicitly tell the assistant "remember this" and manage a flat list of saved facts. Dreaming V3 flips that model entirely. After each conversation ends, the system reviews a user's chat history in the background and automatically decides what is worth remembering, without any manual curation required.

The technical difference matters for both users and OpenAI's infrastructure. Earlier memory systems searched a static list of saved facts at the start of each new conversation, which was computationally expensive when reviewing months of chat history. Dreaming V3 instead runs its synthesis work offline, after conversations finish, rather than during live chats. This separation is why OpenAI can claim the roughly fivefold efficiency gain; the expensive work of reviewing chat history happens once in the background instead of being repeated at the start of every session.

How Does Dreaming V3 Actually Remember You?

The system now behaves more like how a person forms an impression of someone over time than like a traditional database lookup. Instead of simply transcribing what a user says, Dreaming V3 synthesizes information, drawing conclusions across many conversations that a user may never have stated outright in any single chat. For example, if someone mentions being vegetarian across multiple conversations, the system infers and stores that preference without being explicitly told to remember it.

OpenAI paired the more autonomous backend with visible controls on the frontend. Users can now access a memory summary page that shows what the system has pieced together, organized by topic, with timestamps for when each item was last updated. From that page, users can mark an inference as incorrect, tell the model to stop bringing up a topic, or remove an item entirely. Inside individual chats, a "don't mention this again" control lets users suppress a specific detail on the spot.

The system also tracks time in ways the old memory system did not. A note like "planning a trip to Singapore in July" automatically revises to "traveled to Singapore in July 2026" once the date has passed, rather than sitting in the profile as stale information. When ChatGPT answers using something it remembers, the response shows which past chat, saved detail, custom instruction, or uploaded file the model drew on, giving users a paper trail the original saved-memories list never provided.

Steps to Manage Your ChatGPT Memory Settings

• Access the Memory Summary Page: Navigate to ChatGPT's settings to view what the system has automatically learned about you, organized by topic with update timestamps.
• Correct or Remove Inferences: Review each inferred detail and mark any as incorrect, delete items you don't want stored, or update information that has changed.
• Control Topic Mentions: Decide which topics ChatGPT should bring up unprompted in future conversations and which it should leave alone entirely.
• Use In-Chat Suppression: During conversations, use the "don't mention this again" control to immediately suppress specific details without editing your full memory profile.

Why Does This Matter for AI Competition?

The timing of Dreaming V3's launch reflects broader competitive pressures in the AI assistant market. ChatGPT's grip on the AI assistant market has loosened this year as rivals gain ground, and memory, the ability of an assistant to carry context between sessions without being re-told, has become one of the few product features that can lock a user into a single platform. Dreaming V3 is OpenAI's clearest bet yet that remembering a user well is worth more than any single model upgrade.

The 82.8% recall accuracy represents a significant jump from earlier versions. While OpenAI has not published a full benchmark table alongside the launch, early coverage of the release converges on this figure. Reported starting points vary depending on which prior version is used as the baseline, with some accounts citing a jump from roughly 68% accuracy, but the 82.8% endpoint is consistent across multiple independent write-ups of OpenAI's internal evaluation.

When Will Everyone Get This Update?

The rollout is happening in stages. Dreaming V3 became available on June 4, 2026, to ChatGPT Plus and Pro subscribers in the United States first. As of the time of publication, OpenAI has not given a specific date for when Team and Enterprise accounts will receive the update. This staged approach matches how OpenAI has handled every previous memory expansion; the original memory feature launched to a small slice of users in February 2024 before reaching the Free, Plus, Team, and Enterprise tiers by September 2024.

If Dreaming V3 follows the same pattern, the widest rollout to all user tiers is likely still months away. A workplace deployment raises additional questions that individual Plus subscribers do not face: who can see what the model has inferred about an employee, whether HR- or legal-sensitive conversations should be excluded from synthesis by default, and how memory is handled once someone leaves the company. OpenAI is likely to treat those questions as a prerequisite for the Enterprise rollout rather than an afterthought.

What Does This Evolution Tell Us About AI Assistants?

ChatGPT memory has moved through five distinct phases since OpenAI first introduced it in February 2024. Each update handed the assistant a little more initiative to decide, on its own, what was worth keeping. The direction of travel is consistent: every update shifts more of the remembering from the user to the model itself. Dreaming V3 is the point where that shift becomes close to total. The user no longer curates a list; the model curates an evolving model of the user.

For a company running one of the highest-traffic consumer AI products in the world, shaving compute cost per active user by a factor of five is not a minor engineering footnote. The efficiency gains mean OpenAI can offer more sophisticated memory features without proportionally increasing server costs, a critical advantage as the company competes for user loyalty in an increasingly crowded market.