OpenAI replaced the default ChatGPT model on May 5, swapping out GPT-5.3 Instant for the new GPT-5.5 Instant across all user tiers. The headline number is a 52.5 percent reduction in hallucinated claims on high-stakes prompts covering medicine, law, and finance. That is a meaningful improvement for anyone using ChatGPT in any professional context. The model also pulls back on the excessive emoji usage that had crept into responses, returning to something that feels closer to a tool than a cheerful digital assistant. GPT-5.5 Instant also connects to Gmail history, past conversations, and saved files to deliver more personalized answers.
Over 600 million people use ChatGPT. All of them are now running a more accurate model whether they noticed the update or not. GPT-5.5 Instant replaces GPT-5.3 Instant in the API as chat-latest, with the previous model remaining available for three months before it is retired. Enhanced memory and personalization features are rolling out first to Plus and Pro subscribers on the web before reaching Free, Go, Business, and Enterprise tiers. The update received minimal press attention but represents the kind of incremental improvement that compounds across hundreds of millions of daily interactions. A 52 percent reduction in hallucination on medical and legal queries is not a small number when that query is someone making a real decision about their health or their money.
Keywords: GPT-5.5 Instant ChatGPT, OpenAI model update May 2026, ChatGPT hallucination reduction, AI model accuracy