ChatGPT switches to GPT‑5.5 Instant as default worldwide

ChatGPT’s default model just changed, and that matters more than a flashy new model-picker option ever does. Most people never touch the selector. They open ChatGPT, ask a question, and use whatever is already there. On May 5, OpenAI swapped that default worldwide from GPT-5.3 Instant to GPT-5.5 Instant for logged-in users, which means the product most people experience got quietly but materially different. (openai.com) ### What exactly changed? The simple version is this: the “daily driver” model inside ChatGPT got upgraded. OpenAI says GPT-5.5 Instant is now the default for everyone who signs in, and the company frames it as a smarter, clearer, faster-feeling general model rather than a niche reasoning model for specialists. That makes this less like launching a new trim level(openai.com)eady drive. (openai.com) ### Why does “default” matter so much? Because defaults are the product. A lot of AI launches are technically real but behaviorally invisible — they only matter if a user knows where to click. This one lands differently. OpenAI says Instant is the model used for everyday work and learning, and its own release notes call it the “daily driver” for hundreds of milli(openai.com)nd latency show up at huge scale. (openai.com) ### What is OpenAI promising got better? Three things, basically: fewer made-up claims, tighter answers, and better personalization. OpenAI says GPT-5.5 Instant gives clearer and more concise responses, asks fewer unnecessary follow-up questions, and cuts clutter like overformatting and gratuitous emojis. For paid web users, it can also make better use of context(openai.com)n that context helps. (openai.com) ### How big is the accuracy jump? OpenAI’s headline number is a 52.5% reduction in hallucinated claims versus GPT-5.3 Instant on high-stakes prompts in areas like medicine, law, and finance. The company also says inaccurate claims fell 37.3% on especially difficult conversations that users had previously flagged for factual errors. Those are internal evaluations, (openai.com)e clearest signal of what OpenAI thinks improved most. (openai.com) ### Is this the same as full GPT-5.5? Not quite. GPT-5.5 is the broader model family OpenAI introduced in late April, with GPT-5.5 and GPT-5.5 Pro available in the API. GPT-5.5 Instant is the lighter, lower-latency version tuned for everyday ChatGPT use. Think of it as the fast workhorse, while the bigger variants are there when users or developers want more deliberate reasoning and more control. (openai.com) ### What do paid users still get? More knobs. OpenAI’s help pages say Plus, Pro, and Business users can use the model picker to manually choose GPT-5.5 Instant or GPT-5.5 Thinking, turn automatic switching on or off, and set thinking effort. So the free experience got better, but the premium tiers still hold onto control and deeper reasoning options. (help.o([openai.com)5-in-chatgpt)) ### What changed from the old setup? Until now, GPT-5.3 had been the default baseline in ChatGPT after earlier retirements shuffled older models out of the product. This move effectively makes 5.5 the new center of gravity inside ChatGPT, not just an optional upgrade on the side. That matters because it narrows the gap between what enthusiasts get excited about and what ordinary users actually touch. (help.openai.com) ### So what’s the real takeaway? This is not the kind of launch that screams with a new interface. It’s more important than that. OpenAI just changed the answer quality, tone, and reliability of the version of ChatGPT most people meet by default. If the company’s numbers hold up in day-to-day use, the biggest win won’t be that GPT-5.5 Instant exists — i(help.openai.com) name at all. (openai.com)