File photo
Washington, USA: OpenAI announced the launch of its new virtual model GPT-5.5 Instant in the ChatGPT smart chatbot, with notable improvements in the accuracy of information provided, and a reduction in errors, or what is known as AI hallucinations, which is the problem that leads to the generation of false or fabricated information.
According to OpenAI, GPT-5.5 Instant reduced false statements by 52.5% compared to GPT-5.3 Instant on high-stakes prompts in medicine, law, and finance. Hallucinations have been ChatGPT's most persistent flaw since the beginning.
OpenAI also tested against conversations real users had previously flagged for factual errors. On those, inaccurate claims dropped by 37.3%.
OpenAI presents the new model as more efficient at everyday tasks, including analyzing uploaded images with higher accuracy, determining when to use web searches to provide more useful answers, and providing more concise, focused responses with less use of unnecessary emojis.
GPT-5.5 Instant also pulls more actively from your past chats, saved files, and connected Gmail account to make answers personally relevant, in parallel with similar features in Google’s Gemini bot.
The company is introducing a new feature called "memory sources," which allows users to see what data was used to customize responses, with the ability to modify or delete it when needed, thus enhancing transparency and control over information.
OpenAI has begun rolling out GPT-5.5 Instant gradually to all ChatGPT users starting Wednesday, while keeping GPT-5.3 Instant available as an option for three months before it is finally discontinued.
The enhanced customization features are scheduled to arrive first to Plus and Pro subscribers via the web, and will later be available in mobile apps, and then for the rest of the accounts. The "memory sources" feature has also started reaching users via the web, with plans to launch it soon in the mobile app. (
