OPENAI SAYS ITS NEW CHATGPT MODEL HALLUCINATES 52 PERCENT LESS AND ROLLS IT OUT TO EVERY USER ON EARTH OpenAI made GPT-5.5 Instant the new default model for all ChatGPT users in early May, replacing GPT-5.3 Instant across the entire user base simultaneously. The company says the new model produces 52.5 percent fewer hallucinated claims on high-stakes prompts covering medicine, law, and finance compared to its predecessor. That is a significant number if accurate, because hallucination has been the persistent credibility problem preventing AI from being trusted for anything with real consequences attached. The model is also more concise. OpenAI reports it uses 30 percent fewer words and nearly 30 percent fewer lines than its predecessor without losing substance. For users who found earlier ChatGPT responses padded with qualifications and caveats that restated the obvious, this is a meaningful change in day to day usability. For Plus and Pro users on web, the model can reference past conversations, uploaded files, and connected Gmail accounts to personalize answers. A user who discussed a business strategy over several sessions weeks ago can ask a follow-up question and get a response that reflects what was said before rather than starting from scratch. GPT-5.5 Instant went live to all free-tier users on May 5. Two days later, OpenAI launched GPT-5.5-Cyber in limited preview for vetted cybersecurity teams under its Trusted Access for Cyber program, a specialized variant tuned for security research tasks. OpenAI describes the broader arc as building toward a super app experience that can handle everything across a person’s digital life. This upgrade is one more component in that architecture. Keywords: GPT-5.5 Instant, ChatGPT new model, OpenAI hallucination, GPT-5.5 update