**[OpenAI Launches Next-Generation ChatGPT Memory System "Dreaming" to Optimize Freshness, Continuity, and Relevance]** BlockBeats News, June 5 — OpenAI announced the rollout of a more powerful and scalable ChatGPT memory synthesis system called "Dreaming," designed to enhance the freshness, continuity, and relevance of memory while addressing issues of staleness, accuracy, and scalability when applying memory across hundreds of millions of users and over extended time spans. OpenAI stated that the memory feature enables ChatGPT to learn user preferences, projects, and constraints, allowing future conversations to build on shared context rather than starting from scratch. This update is available starting today for Plus and Pro users in the U.S. and will expand to more countries and Free and Go users in the coming weeks. According to the announcement, ChatGPT first introduced the "Save Memory" feature in April 2024, allowing users to request ChatGPT to remember specific information. In April 2025, OpenAI launched the first version of Dreaming, enabling ChatGPT to reference chat history through backend processes and automatically organize and synthesize memory states. The newly released Dreaming V3 is a next-generation memory architecture built on this mechanism. OpenAI claims the new system improves factual recall, adherence to preferences, and accuracy over time. Evaluations show that the success rate for factual recall tasks increased from 41.5% in 2024 to 82.8% in 2026; the success rate for preference adherence rose from 31.4% to 71.3%; and the success rate for maintaining accuracy over time improved from 9.4% to 75.1%. OpenAI also noted that recent optimizations have reduced the computational power required to provide Dreaming to Free users by approximately fivefold, paving the way for its rollout to Free users in the coming weeks while also increasing memory capacity for Plus and Pro users. [Original Link]