OpenAI Starts Rolling Out New GPT-5.5 Instant Model and Memory Sources to ChatGPT Users

OpenAI released GPT-5.5 Instant yesterday, its new default model for ChatGPT that will replace the GPT-5.3 Instant model that it shipped back in March. GPT-5.5 Instant should provide more accurate answers with fewer hallucinations, and for Plus and Pro users, the new model will also make better use of context from past chats and connected services like Gmail to improve its responses.

OpenAI’s “Instant” GPT models are optimized for low-latency and efficiency, which is why they’re used by default by the chatbot. For more complex tasks, ChatGPT may automatically switch to the GPT-5.5 Thinking model and use deeper reasoning before answering. This “Thinking” mode can also be manually selected in ChatGPT.

“GPT-5.5 Instant is more factually reliable, especially for prompts where accuracy matters most. It also gives tighter, more direct answers without losing useful details, asks fewer unnecessary follow-up questions, and reduces clutter like overformatting and gratuitous emojis,” OpenAI explained in the release notes for GPT-5.5.

As I mentioned earlier, ChatGPT Plus and Pro users will get more relevant responses when using GPT-5.5 Instant as the new model better leverages information from past chats, saved memories, and Gmail inboxes (when they’re connected). OpenAI is also rolling out memory sources to all ChatGPT users, which will show what information the chatbot used to personalize its responses.

On the web, memory sources can be accessed by clicking the ‘Sources’ icon below a specific response. ChatGPT users can also choose to make corrections or delete saved memories from the chatbot’s side panel. Memory sources will be rolling out soon to mobile users, OpenAI said yesterday.

For ChatGPT paid users, enhanced personalization is currently available on the web and coming soon to mobile. The old GPT-5.3 model will also remain available for 3 months via ChatGPT’s model picker.