OpenAI is rolling out improved memory for ChatGPT to all users

If you’ve been using ChatGPT for a while, you’d know that its memory feature can be quite useful. It remembers important information about you and tailors responses based on what it already knows. Now, OpenAI is making even more improvements to the feature for all users and bringing its “Dreaming” memory system to users on the Free plan as well.

The company today announced that it’s introducing a new memory architecture to ChatGPT which will make the chatbot much better at remembering, synthesizing, and carrying forward context.

It’s based on ChatGPT’s existing “Dreaming” process, which allowed the AI chatbot to curate information about users in the background using existing chat threads.

The new memory system introduces several important changes: it will provide users with a readable summary of what ChatGPT knows about them, allow them to add or update information, and tell ChatGPT which topics it should bring up and when.

It also helps keep the AI’s memory updated. OpenAI offers an example of a user planning a trip to Singapore. Normally, ChatGPT would remember that the user is traveling, but it was unlikely to understand when the trip ended and could provide answers under the assumption that the user was still in Singapore.

With the new architecture, memories are automatically updated as time passes. So, ChatGPT can “revise its memory from ‘You’re going to Singapore in July’ to ‘You went to Singapore in July 2026’ when the trip ends,” the company says.

OpenAI also claims that this new memory architecture is not only more capable, but also more compute-efficient. It’s because of this increased efficiency that the company is now planning to roll out the improved memory feature to Free users as well.

The new memory architecture is rolling out to Pro and Plus users in the US starting today. It will expand to other countries and will be offered to Free and Go users in the coming weeks as well.