OpenAI released GPT-5.3 Instant, the most used model update for ChatGPT, focusing on making daily conversations more practical and smoother. Officials say that this version reduces unnecessary refusals and redundant disclaimers, and provides richer and more contextually appropriate answers in online search scenarios, while reducing mechanization and blunt tone.
The update also emphasizes reliability improvements: in internal evaluations, GPT-5.3 Instant shows a decrease in hallucination rates in high-risk fields such as healthcare, law, finance, and high-incidence scenarios. For developers, the API can use a model called gpt-5.3-chat-latest; corresponding updates for Thinking and Pro versions will be released one after another.
The official also reminds that there are still limitations: some non-English languages (such as Japanese and Korean) may have a sense of literal translation, and the naturalness of intonation continues to be optimized. GPT-5.2 Instant will be reserved for three months for paying users in the form of "Legacy Models" and is scheduled to be retired on June 3, 2026.
FAQs
Q: What are the main experience points changed in GPT-5.3 Instant?
A: The focus is on reducing unnecessary rejections and disclaimers, improving tone fluency, and strengthening the integration quality of online search answers.
Q: When will GPT-5.3 Instant be available?
A: Officials say that it will be open to all ChatGPT users from the release date, and developers can also call the corresponding model through the API.
Q: Which model name should developers use in the API?
A: The official API model is called gpt-5.3-chat-latest.
Q: How long will GPT-5.2 Instant be retained?
A: GPT-5.2 Instant will be reserved for three months in the traditional model menu for paying users, and is scheduled to be retired on June 3, 2026.
Q: What are the known limitations of GPT-5.3 Instant?
A: Some non-English languages may be more stiff or have a sense of literal translation, and the tone and customization options are still being iterated.
