OpenAI asserts that the latest ChatGPT model significantly reduces hallucinations.

OpenAI’s latest model for ChatGPT, GPT-5.5 Instant, aims to tackle the issue of inaccuracies that have plagued AI models. The company asserts that this new version has made “significant improvements in factuality across the board.” According to internal evaluations, GPT-5.5 Instant generated “52.5% fewer hallucinated claims” compared to the previous GPT-5.3 Instant model, particularly on high-stakes topics like medicine, law, and finance. Additionally, it has cut down on inaccurate claims by 37.3% in challenging conversations that users identified as having factual errors.

Beyond accuracy, OpenAI highlights that GPT-5.5 Instant is “more capable across everyday tasks,” such as analyzing image uploads and knowing when to seek information online. The responses are described as “tighter and more to-the-point,” with a reduction in unnecessary emojis.

With this update, ChatGPT is now “more effective” at incorporating context from previous chats and your Gmail, allowing for more personalized interactions. A new “memory sources” feature will also enable the chatbot to display the context used for these personalized responses, giving users the option to delete or correct information as needed.

OpenAI plans to roll out GPT-5.5 Instant to all ChatGPT users starting Tuesday, while the GPT-5.3 Instant model will remain available for three months before being phased out. This transition period is designed to ease the shift for users who may be attached to older models. Enhanced personalization features will initially be available to Plus and Pro users on the web, with mobile app updates coming soon. OpenAI also intends to extend these features to Free, Go, Business, and Enterprise users in the near future. The memory sources feature is currently being rolled out to ChatGPT consumer plans on the web and will soon be available on mobile.