ChatGPT’s default model has been upgraded to GPT-5.3 Instant to smooth conversationsOpenAI promises this will “reduce the cringe” of caveats and refusalsChatGPT 5.3 should also have fewer hallucinations
OpenAI has heard the complaints that ChatGPT too often answers a simple question with a long list of its limitations, and has released GPT-5.3 Instant as the new default model for ChatGPT to, in its words, “reduce the cringe” of those rough conversations. The upgrade is designed to reduce unnecessary refusals and shorten safety disclaimers to improve the AI’s overall tone.
Earlier versions of the system sometimes took a defensive approach when answering questions. Even when a request was harmless, the model might begin with a lengthy explanation of safety rules before finally addressing the prompt. In other cases, it might decline to answer a question entirely because it misinterpreted the intent behind it.
GPT-5.3 Instant tries to fix that behavior. Instead of leading with warnings or hypothetical misuse scenarios, the system now prioritizes delivering a direct answer when the question is reasonable. Safety restrictions remain in place, but the model has been tuned to avoid unnecessary caution.
You may like
The upgrade will simply appear the next time most people open the chatbot and should boost ChatGPT’s intellect as well as its social niceties. It’s supposed to upgrade the more subtle qualities that determine whether talking to an AI feels pleasant or annoying — the “cringe,” if you will.
We heard your feedback loud and clear, and 5.3 Instant reduces the cringe. pic.twitter.com/WqO0XzLcVuMarch 3, 2026
Now, if I were young and hip, saying something “reduces cringe” would only make me think the cringe factor was even higher. But if the new model means ChatGPT is less stiff and doesn’t deploy a firehose of caveats with every prompt, I’ll accept the slightly desperate wording OpenAI uses.
There should be less to warn about, too, if it has the improved accuracy promised by OpenAI. ChatGPT-5.3 Instant should have fewer hallucinations — roughly 27% fewer when doing online research, though only 20% fewer without going online to check. It should also be better at blending web data with its existing knowledge.
GPT-5.3 personality
Replacing ChatGPT 5.2 with ChatGPT 5.3 changes the baseline experience across millions of daily chats. The update touches everything from homework help to recipe ideas to casual curiosity about how something works. GPT-5.2 Instant will not disappear entirely. Paid users will still have access to it as a legacy model.
The emphasis on social abilities is a notable evolution from when every AI developer boasted about solving the most difficult problems or performing the most complicated tasks. The focus now is on a product that’s pleasant to use — and that means making AI pleasant to talk to.
A chatbot that sounds overly cautious is irritating, and if you want people to keep coming back, even subtle design decisions about response style can be crucial. The winner of the AI chatbot battle may not be the flashiest or most powerful model, but just one that doesn’t remind people of an annoying pedant who takes five minutes to answer a question when a single sentence will do.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
The best business laptops for all budgets
Our top picks, based on real-world testing and comparisons