GPT-5.5 Instant: 3 changes ChatGPT users will notice right away
OpenAI quietly swapped in a new default ChatGPT model on May 5, with 52 percent fewer made-up facts and 30 percent shorter responses.
Noticed lately that ChatGPT gets to the point instead of writing a three-paragraph essay? It is not random.
OpenAI swapped the default model on May 5 with something it calls GPT-5.5 Instant internally. The transition happened quietly, with no in-app announcement. Users noticed first that responses were getting shorter and at the same time more precise.
GPT-5.5 Instant replaced the previous GPT-5.3 Instant as the default on every ChatGPT plan, from Free to Enterprise. The change does not require any action; the model takes over when you start a new conversation.
1. Fewer made-up facts in high-stakes domains
OpenAI published its own benchmark results. Treat these with caution (the company is grading its own product), but the numbers stand out.
On questions in medicine, law and finance, the new model produces 52.5 percent fewer fabricated claims (commonly called “hallucinations,” where the model writes something that sounds credible but does not exist in reality) than the previous version.
This matters for users who bring real questions to the model. The earlier version was known to invent drug names or legal articles that do not exist.
In practice, if you ask about a medication dose or a statute of limitations, your chances of getting a correct answer are higher. But the rule still holds: for anything important, check with a professional.
2. About thirty percent shorter answers
The new model uses 30.2 percent fewer words and 29.2 percent fewer lines in a typical response.
The old ChatGPT had a habit of opening with phrases like “that’s a great question” before reaching the point. The new model usually skips that and goes straight to the answer.
If you use the model to write emails or quick notes, shorter responses mean less editing before sending. The difference is noticeable after a few days of use.
3. Memory across past chats and connected services
The biggest leap in personalization. ChatGPT can now pull context from:
- Past conversations
- Files you have shared with it
- Connected Gmail, if you have allowed it in settings
Plus and Pro users on the web already have this option. The mobile version is arriving shortly. Free and Business plans will follow in the coming weeks.
Memory is not on by default. It has to be enabled manually under Memory and personalization settings. If you are privacy-conscious, you do not need to turn it on at all. The model still works as before.
What this means for an everyday user
If you use ChatGPT for work, the biggest gain is response speed and precision. Less time spent rewriting what the model gave you.
If you use it for personal finance advice or to make sense of medical reports, the new model is more reliable. It is still not a replacement for a professional.
OpenAI is also opening ChatGPT for Excel and Google Sheets globally, as a sidebar integration. Spreadsheets can now be built and edited directly from ChatGPT.
What users in Serbia need to know
Serbian payment cards work for ChatGPT Plus at 20 dollars per month (about 2,350 dinars) and the Pro plan at 200 dollars (about 23,500 dinars).
There are no geographic restrictions on access to the new model. The handover is automatic.
ChatGPT for Excel works if you have a Microsoft 365 subscription, which is supported normally in Serbia.
Have you noticed ChatGPT responses getting shorter over the past few weeks?
