ChatGPT's new default model is more factual and better at personalization

Hot on the heels of its recent GPT-5.5 release, OpenAI is updating ChatGPT's default model to GPT-5.5 Instant. OpenAI touts GPT-5.5 as producing 52.5 percent fewer hallucinated claims in internal testing, and 37.3 percent fewer inaccurate claims when taking on conversations ChatGPT users had flagged for factual errors.

At the same time, OpenAI says it designed the new model to generate more direct responses, "while keeping the warmth and personality that makes ChatGPT enjoyable to use." GPT-5.5 Instant is also allegedly better at asking fewer unnecessary followup questions and should generate less cluttered outputs, in part thanks to more subdued emoji use. Separately, OpenAI has made the new model faster at searching uploaded files, connected Gmail accounts and prior ChatGPT conversations to generate personalized responses. The upgrade should translate to users needing to repeat themselves less often.

Alongside GPT-5.5 Instant, OpenAI is introducing a feature it calls memory sources, which is designed to give users more visibility over the context ChatGPT may have used to personalize a response. It also gives people the option to delete the information, or update it if it's no longer relevant.

"Memory sources aren't shown to others if you choose to share a chat," OpenAI explains. "You remain in control of what's in your memory: you can delete chats you no longer want to be cited, delete or change items in saved memories in settings, or use temporary chats that don't use or update your memory."