San Francisco: OpenAI has recently rolled out GPT‑5.3 Instant, a better version of its previous Instant model, to make its everyday conversations more clear, factually correct, and natural for users.
Earlier versions of ChatGPT used to sometimes avoid answering even safe, simple questions. They would add long warnings or cautious messages before giving the real answer. GPT‑5.3 Instant mends this. Now, if the question looks harmless, the model answers it directly, raising no complaints, no demands.
The new update also changes how ChatGPT makes use of the information it gathers from the web. Instead of simply listing links or giving random summaries, the model now fuses online information with its own knowledge to give clear answers. It puts in great efforts to understand what the user really wants and presents the most important details first to make the responses easier to follow.
Tone is another crucial area of improvement. In the past, ChatGPT used to sound a little too formal or assumed the user was upset or stressed on its own. GPT‑5.3 Instant now replies in a more natural style. It avoids awkward wording and unnecessary emotional comments, making conversations feel smooth. Users can also adjust the tone in settings if they prefer a different style.
OpenAI also reports that the new model is more factually reliable. Tests show that it makes far fewer mistakes now. Statistically speaking, it makes 26.8% fewer mistakes with web info and 19.7% fewer using just its own knowledge. This makes its answers more trustworthy across topics like medicine, finance, and law.
GPT‑5.3 Instant is now available for all ChatGPT users, while the previous model is set to retire on June 3, 2026.
Previous
