OpenAI’s ChatGPT gets memory update, teases new AI model

OpenAI is upgrading ChatGPT’s memory so that the AI chatbot can now recall previous conversations and customise responses on their basis. The feature will be rolled out to ChatGPT Plus and Pro users at the moment and gradually to Enterprise, Team and Edu users over the next weeks. 

It’s still not known if the update will be available to basic ChatGPT users yet. 

CEO Sam Altman announced the update on X on April 10 saying, “This is a surprisingly great feature imo, and it points at something we are excited about: AI systems that get to know you over your life, and become extremely useful and personalised.” 

While OpenAI had included a feature called ‘Memory’ earlier, users had to turn it on to specifically ask ChatGPT to remember something for later. The feature could only retain a limited amount of information for the future. 

The default option for chats was the ‘Temporary Chat’ option. 

Now, ChatGPT will have two new options - ‘Reference saved memories,’ which is similar to old feature and ‘Reference chat history’ which asks the chatbot to apply context from previous conversations and adapt future conversations.

Users will still be given the option to disable the feature completely. 

Separately, on the same date, Altman also teased on X that they will soon be launching an exciting feature today. Although there are rumours that this is potentially a new large language model, it’s unclear whether it will be the successor to GPT-4 or a new reasoning AI model. 

A new report by The Verge has stated that the outlet is expecting OpenAI to release GPT-4.1, alongwith a mini and nano version by next week. Sources also found references to o3 and o4 models in a new ChatGPT web version which could indicate that their launches are also looming ahead.