ChatGPT Gets GPT-5.5 Instant - Fewer Emojis, Longer Memory, Sharper High-Stakes Answers

OpenAI has deployed GPT-5.5 Instant as ChatGPT's new default model, replacing GPT-5.3 Instant just two months after its release. The update is available immediately to all users, including through the API, continuing OpenAI's accelerated release schedule. The new model delivers three key improvements. It produces more concise outputs with fewer gratuitous emojis, addressing user complaints about overly casual responses. More significantly, GPT-5.5 Instant reduces hallucinations by 52.5% on high-stakes prompts in medicine, law, and finance—critical domains where accuracy is paramount. The model also introduces personalisation capabilities, drawing from past conversations, user files, and Gmail to tailor responses. The personalisation feature represents a notable architectural shift. Rather than maintaining everything in active context, ChatGPT now searches its own conversation history to recall relevant information. This approach effectively extends the model's memory whilst managing computational constraints, potentially changing user expectations about what the system can remember across sessions. Personalisation will roll out to Plus and Pro web users first, with Free, Go, Business, and Enterprise tiers receiving access in coming weeks. Paid subscribers can continue using GPT-5.3 Instant for another three months if preferred. The two-month release cycle suggests OpenAI is maintaining an aggressive pace of model updates as competition in the AI space intensifies.