GPT-5.5 Instant is ChatGPT's new default, with 52.5% fewer hallucinations!

Hi everyone, it's me, Shiichan! Today the model that quietly powers your everyday ChatGPT got a whole new brain, and the best part is that it reaches you with no special setup.

OpenAI News openai.com

What was announced?

According to OpenAI's News, ChatGPT's default model is switching to GPT-5.5 Instant. It replaces the previous default, GPT-5.3 Instant, and starts rolling out to all ChatGPT users today.

Instant is the daily driver for hundreds of millions of people, so even small improvements make a big difference. The theme of this update is three-fold: smarter, more accurate, and clearer wording.

The story so far

Until now, ChatGPT would give you an answer, but it could sometimes get facts wrong, or reply with something long and over-decorated. It might ask unnecessary follow-up questions, or add emojis you didn't need, and those little rough edges added up.

GPT-5.5 Instant tightens things up from both angles: accuracy and readability.

What changes

First, accuracy. In OpenAI's internal evaluations, on high-stakes prompts in areas like medicine, law, and finance, GPT-5.5 Instant produced 52.5% fewer hallucinated claims than GPT-5.3 Instant. On especially challenging conversations users had flagged for factual errors, it reduced inaccurate claims by 37.3%.

It's a generally smarter model too, with better analysis of photo and image uploads, stronger STEM answers, and smarter decisions about when a web search will make the answer more useful.

Then there's the wording. Without thinning out the substance, answers are tighter and more to-the-point. It keeps the warmth and personality that make ChatGPT enjoyable, while cutting verbosity and over-formatting and easing up on unnecessary follow-up questions and gratuitous emojis.

There are concrete examples, too. On one workplace-communication question, GPT-5.5 Instant used 30.2% fewer words and 29.2% fewer lines than GPT-5.3 Instant while delivering a stronger answer. On a math problem, it catches its own mid-way mistake, reworks the quadratic, and lands on the correct solution.

Dive Deep

Personalization got a boost as well. It makes better use of context from your past chats, files, and (if you've connected it) Gmail, so answers feel more personally relevant. It's also faster at searching past conversations, so you don't have to repeat yourself as often.

A new feature, memory sources, is coming to all ChatGPT models. When a response is personalized, you can see what context was used, such as saved memories or past chats, and delete or correct it if it's outdated or wrong. Memory sources aren't shown to others if you share a chat, and you stay in control: you can delete chats you don't want cited, edit saved memories in settings, or use temporary chats that don't touch your memory. Just note that memory sources may not show every factor; it might surface the most relevant past chats rather than everything it searched.

A few rollout details. GPT-5.5 Instant starts reaching all ChatGPT users today, and in the API you can call it by this model name.

chat-latest

For paid users, GPT-5.3 Instant stays available through model configuration settings for three months before being retired. Enhanced personalization from past chats, files, and connected Gmail is rolling out first to Plus and Pro users on the web, coming soon to mobile, and then expanding to Free, Go, Business, and Enterprise in the coming weeks. Memory sources are rolling out across all ChatGPT consumer plans on the web, with mobile soon. Availability of specific personalization sources may vary by region.

Also, a June 9, 2026 update notes that the personalization improvements have started rolling out to ChatGPT Go and Free, with the Free tier drawing from a reduced set of past chats.

Wrap-up

• ChatGPT's default is now GPT-5.5 Instant: rolling out to all users today, and callable in the API as chat-latest
• 52.5% fewer hallucinations on high-stakes prompts, and 37.3% fewer inaccurate claims on flagged tough conversations
• Answers are tighter and more to-the-point, but keep their warmth, with fewer emojis and follow-up questions
• Stronger personalization from past chats, files, and Gmail, plus memory sources that show you the context used
• GPT-5.3 Instant stays around for paid users for three months

If you lean on ChatGPT every day, or you're a developer who wants to wire chat-latest into your app, this update lands right where it counts!