OpenAI has replaced GPT-5.3 Instant as ChatGPT's default model with GPT-5.5 Instant, a significant update that the company says delivers its most factually reliable everyday AI to date. The rollout, which began on May 6, 2026, affects hundreds of millions of users worldwide and marks the most consequential update to ChatGPT's default experience since GPT-5 launched earlier this year.

"GPT-5.5 Instant produced 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts covering medicine, law, and finance."

— OpenAI

The headline improvement is factual accuracy. In internal evaluations, GPT-5.5 Instant produced 52.5% fewer hallucinated claims than its predecessor on high-stakes prompts covering areas like medicine, law, and finance. It also reduced inaccurate claims by 37.3% on especially challenging conversations that users had previously flagged for factual errors. For a model that serves as the daily interface for hundreds of millions of people, these are not incremental gains — they represent a meaningful reduction in the risk of consequential misinformation.

Tighter Answers, Better Reasoning

Beyond accuracy, GPT-5.5 Instant is designed to be more concise without sacrificing substance. OpenAI says the model delivers the same information as its predecessor, often with more utility, while reducing the verbosity and overformatting that made previous responses feel cluttered. It asks fewer unnecessary follow-up questions, avoids gratuitous emojis, and produces tighter prose across a wider range of tasks. Benchmark improvements are visible across visual reasoning, mathematics, and science.

The model also shows meaningful improvements in analyzing photo and image uploads, answering STEM-related questions, and deciding when to use web search to provide a more useful answer. These gains reflect a broader shift in how OpenAI is approaching model development: rather than chasing benchmark records on narrow tasks, the company is optimizing for the quality of everyday interactions at scale.

Memory Sources: A New Layer of Transparency

Perhaps the most structurally significant change in GPT-5.5 Instant is the introduction of memory sources — a new feature that gives users visibility into what context was used to personalize their responses. When a response is personalized using saved memories, past chats, or connected Gmail data, users can now see exactly what information was referenced, and delete or correct it if something is outdated or no longer relevant.

This represents a meaningful step toward AI transparency. Previous versions of ChatGPT's memory system operated as a black box: users knew that the model could remember things, but had limited visibility into what it actually remembered or when that memory was being applied. Memory sources change that dynamic by surfacing the specific context that shaped a given response, giving users a degree of oversight that has been largely absent from consumer AI products.

Enhanced personalization from past chats, files, and connected Gmail is rolling out to Plus and Pro users on the web first, with mobile support and expansion to Free, Go, Business, and Enterprise tiers planned for the coming weeks. Memory sources are rolling out across all ChatGPT consumer plans on the web, with mobile support to follow.

API Access and Model Continuity

For developers, GPT-5.5 Instant is now available via the API as the model identifier chat-latest. GPT-5.3 Instant will remain available for paid users for three months through model configuration settings before being retired, giving developers time to test and migrate their applications. The launch comes as OpenAI faces intensifying competition from Anthropic's Claude and Google's Gemini, both of which have posted strong user growth numbers in recent months.