OpenAI recently released GPT- 5.5 Instant and made it the new default model for ChatGPT, replacing GPT-5.3 Instant. No pop-up alert. No countdown timer. Just a switch.
GPT-5.5 Instant is not a completely new model family. It sits within the GPT-5 generation, a series OpenAI began rolling out in mid 2025. It appears to be an iterative evolution within the GPT-5 series rather than a wholly separate model family.
The new model is rolling out to all ChatGPT users, replacing GPT-5.3 Instant as the default. The clearest improvement is in accuracy. Reportedly, in internal evaluations, GPT-5.5 Instant produced 52.5% fewer hallucinated claims that GPT-5.3 Instant on high-stakes prompts covering areas like medicine, law, and finance. And, that is a significant enhancement for users who rely on ChatGPT for professional research or factual lookups.
On the AIME 2025 math test, the new model scored 81.2, compared to 65.4 for the older model. According to reports, it also outperformed its predecessor on the MMMU-Pro multimodal reasoning benchmark, scoring 76 versus 69.2.
The personalised push
Beyond accuracy, OpenAI has made enhancements in memory and personalisation with this update.
GPT-5.5 Instant can now use its search tool to refer back to past conversations, files, and Gmail to give more personalised and curated answers. These features are available to Plus and Pro users on the web, with plans to roll it out to mobile soon.
It matters to users because most AI chatbots treat every conversation like it’s the first one. This new model aims to break that pattern. If you told ChatGPT last month that you live in Mumbai or any other particular place, the model can now factor that in when you ask for restaurants, places to visit kind of suggestions, without you repeating yourself.
OpenAI is also introducing memory sources across all ChatGPT models, giving users visibility over what context was used to personalise responses. Users can delete outdated sources or correct them if the answer was wrong, and memory sources are not shown to others if a chat is shared.
What happened to GPT-5.3?
GPT-5.3 does not disappear immediately, but it is on a clock. For paid users, GPT-5.3 Instant remains available for three months, accessible through model configuration settings, before being retired.
For developers, the GPT-5.5 model is available through the API as “chat-latest”, with GPT-5.3 available as an option for paid users for only three months.
Free users do not have such option. When OpenAI changes the default, free users are going to get what they are given.
How did ChatGPT get here?
To understand this update, it helps to trace the GPT-5 family’s evolution over the past year.
OpenAI launched GPT-5 in August 2025, according to reports GPT-5’s responses with web search enabled were around 45% less likely to contain a factual error that GPT-4o. With thinking enabled, that figure rose to approximately 80% fewer factual errors than OpenAI o3.
GPT-5.1 followed in November 2025, improving response speed and reasoning. GPT-5.1 Instant introduced adaptive reasoning, deciding when to think before responding to more challenging questions, resulting in more thorough answers while still responding quickly.
GPT-5.2 arrived in December 2025, focusing on reliability and reducing errors in complex domains. GPT-5.3 followed in early 2026 and was itself an improvement in conversational tone and web search quality. GPT-5.3 Instant focused on reducing unnecessary dead ends, caveats, and overly declarative phrasing that can interrupt the flow of conversation.
GPT-5.4 then arrived with a focus on coding and agentic tasks. GPT-5.4 brought together advances in reasoning, coding, and agentic workflows, incorporating the coding capabilities of GPT-5.3-Codex while improving how the model works across software environments and professional tasks.
Now GPT-5.5 Instant takes the default slot.
GPT-5.5 beyond “Instant”
There is a separate, more powerful version of GPT-5.5 that was released weeks before the Instant upgrade.
GPT-5.5 rolled out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with GPT-5.5 Pro also rolling out to Pro, Business, and Enterprise users in ChatGPT.
The benchmark numbers for GPT-5.5 proper are significant. On GDPval, which tests agents’ abilities to produce well-specified knowledge work across 44 occupations, GPT-5.5 scores 84.9%. On OSWorld-Verified, which measures whether a model can operate real computer environments on its own, it reaches 78.7%.
The practical use cases are real. Derya Unutmaz, an immunology professor at the Jackson Laboratory for Genomic Medicine used GPT-5.5 Pro to analyse a gene-expression dataset with 62 samples and nearly 28,000 genes, producing a detailed research report that he said would have taken his team months.
It signals where OpenAI is pushing the product: not just chat, but deep professional work.
Does the backlash problem still exist?
OpenAI has a pattern worth noting. It retires old models, users push back, and the company sometimes adjusts. When OpenAI withdrew its GPT-4o model, there was significant backlash from users who related to the model’s personality. Despite the outcry, GPT-4o was deprecated in February 2026.
The emotional attachment users form with specific model versions is a real design problem. People don’t just use these tools — some genuinely feel they “know” a particular model’s style, tone, and quirks. When that changes overnight, it feels personal.
OpenAI addressed this directly when GPT-5.1 launched. The company said that going forward, when new ChatGPT models are introduced, the approach is to give people ample space to evaluate what’s changed and share feedback, and that sunset periods will be communicated clearly and with plenty of advance notice.
Whether GPT-5.5 Instant’s rollout reflects that promise is debatable. Users got the switch on May 5. The three-month window for paid users to revert to GPT-5.3 exists, but free users don’t get that grace period at all.