OpenAI launches ChatGPT-5.5 Instant as default model with better accuracy & memory sources

OpenAI has released GPT-5.5 Instant as the new default model for all ChatGPT users, replacing GPT-5.3 Instant in both the web interface and API. This update brings clearer, more concise, and better personalized answers, as the model now uses previous chats, files, and connected Gmail when relevant, while prioritizing user control.

The model’s dependability and accuracy have increased, achieving 52.5% fewer hallucinated claims and 37.3% fewer inaccurate answers on flagged conversations compared to GPT-5.3 Instant, especially in high-stakes fields like medicine, law, and finance. Everyday interactions benefit from a more natural conversational tone and targeted, useful responses, with enhanced performance in image analysis, science, technology, engineering, and math (STEM) queries, and choosing when to use web search.

These changes enable more concise responses, reduced verbosity, and fewer unnecessary follow-up questions, improving overall readability. In addition, users now see memory sources, a feature giving transparency and control over the context used in personalized responses, rolling out across all ChatGPT consumer plans on the web and coming soon to mobile. While paid users retain access to GPT-5.3 Instant for three months, broader personalization features and memory are set to expand to more user plans in the coming weeks.