OpenAI replaces ChatGPT default model with GPT-5.5 Instant

OpenAI has replaced ChatGPT's default model with GPT-5.5 Instant. The update is rolling out to all ChatGPT users and to the API as chat-latest.

The new model succeeds GPT-5.3 Instant as the version most users will encounter by default. Paid users will continue to have access to GPT-5.3 Instant for three months through model settings before it is withdrawn.

OpenAI said the change is intended to make ChatGPT more accurate, more concise, and better able to tailor replies to individual users. It said the model has been updated across a broad range of everyday tasks, including image analysis, STEM questions, and decisions about when to use web search in a reply.

A central part of the update is reducing false or misleading responses. In internal evaluations, GPT-5.5 Instant produced 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts in fields including medicine, law, and finance, according to OpenAI.

It also reduced inaccurate claims by 37.3% on difficult conversations that users had previously flagged for factual errors. The figures underscore the industry's continued focus on trust and reliability as AI developers try to make consumer tools more useful for everyday work and personal tasks.

Personalisation tools

Alongside the model change, OpenAI is introducing what it calls memory sources across all ChatGPT models. The feature is designed to show users what context was used to personalise a response, including saved memories and past chats, while allowing them to delete or correct outdated information.

OpenAI said GPT-5.5 Instant is better at drawing on context from earlier chats, uploaded files, and connected Gmail accounts when those options are enabled. That should reduce the need for users to repeat information and help with tailored suggestions, ongoing work, and follow-up tasks.

Memory sources will not be visible to other people when a chat is shared, OpenAI said. Users can also delete chats they no longer want cited, change or remove saved memories in settings, or use temporary chats that do not use or update memory.

OpenAI acknowledged that the feature may not show every factor that shaped a response. In some cases, it may display only the most relevant past chats rather than every conversation the system searched or referenced.

Model behaviour

GPT-5.5 Instant is intended to give tighter answers with less unnecessary formatting while keeping a conversational tone, OpenAI said. The model also asks fewer unnecessary follow-up questions and avoids clutter such as excessive emoji use.

That reflects a broader push among AI providers to move beyond headline model launches and focus on how systems behave in day-to-day use. For mainstream chatbot users, changes in brevity, tone, and context handling can be as noticeable as gains in benchmark performance.

OpenAI described Instant as the version used every day by hundreds of millions of people. At that scale, even small shifts in factual accuracy, search behaviour, or personalisation can affect a large base of users across study, work, and routine information-seeking.

OpenAI is staging the wider release of some personalisation features. Enhanced use of past chats, files, and connected Gmail is being introduced first to Plus and Pro users on the web, with mobile support due later, before a broader expansion to Free, Go, Business, and Enterprise plans.

Memory sources are rolling out across all ChatGPT consumer plans on the web, with mobile support to follow. Availability of specific personalisation sources may differ by region.

OpenAI said memory sources are intended to make personalisation easier to understand, while acknowledging that the view is not yet comprehensive. "We'll continue improving this view to make it more comprehensive over time," it said.