OpenAI's enhanced memory now rolling out to free ChatGPT users
5x compute reduction brings dreaming memory to free tier with doubled capacity.
OpenAI announced a major upgrade to ChatGPT's memory feature, introducing a more capable and compute-efficient 'dreaming' system for synthesizing memory. The new architecture addresses challenges of staleness, correctness, and scalability across hundreds of millions of users over multi-year time horizons. Key improvements include better carrying forward of useful context, stricter adherence to user preferences and constraints, and staying current with the passage of time. The system automatically fills in memory gaps and provides a summary page where users can review highlights, add or update information, and instruct ChatGPT on topics to bring up.
Rollout begins today for Plus and Pro users in the US, who also receive doubled memory capacity. Crucially, OpenAI reduced the compute required to serve dreaming to free users by approximately 5x, making it practical to offer the feature at scale. Free users will gain access over the coming weeks. Those who prefer the legacy saved-memories experience can switch back in settings. This marks a significant step toward personalized, continuous AI interactions for a broader audience.
- New 'dreaming' memory architecture uses 5x less compute to serve free users
- Plus/Pro users in the US get upgraded memory and 2x capacity starting today
- Free users will receive the memory feature over the next few weeks
Why It Matters
Personalized AI interactions become accessible to all users, enhancing productivity and continuity across conversations.