Overview

On June 4 2026 OpenAI began rolling out its most capable “dreaming” memory yet — a background process that synthesizes who you are from past chats to optimize for freshness, continuity and relevance, without you ever saying “remember this.” An unofficial bilingual digest of the announcement.

5× less compute to serve Free users (approx.)
3axes of “good memory” OpenAI evaluates
3generations compared: 2024 / 2025 / 2026
2years memory has grown into ChatGPT

How it works

What “dreaming” is, and how a background process turns scattered conversations into fresh, relevant context.

Evolution

How ChatGPT memory grew — straight from OpenAI's own account.

Core capabilities

OpenAI defines “good memory” as doing three jobs well — each shown here with the example from the article.

How it's evaluated

OpenAI evaluates memory on three axes across three generations (2024 saved memories → 2025 + Dreaming V0 → 2026 Dreaming V3) and reports the lift as charts. The article publishes no citable percentages, so this page invents none — it shows the directional claim. The one concrete figure: serving compute cut ~5× for Free users.

Your controls

What the article says you can do, plus general controls — the official Memory FAQ is the source of truth.

Rollout & tiers

Available to Plus and Pro users in the US on June 4 2026; rolling out to additional countries and to Free and Go users over the coming weeks. A ~5× reduction in compute to serve Free users is what makes that practical, alongside increased memory capacity for Plus and Pro.

Privacy & risks

Extended context — general, well-known considerations for a memory that persists separately from your chats. These are NOT part of OpenAI's announcement; for exact controls and policy, rely on the official Memory FAQ and your local settings.

FAQ

Quick answers about ChatGPT's dreaming memory.