OpenAI has introduced a new version of ChatGPT — GPT-5.3 Instant

OpenAI has released a new version of its most popular ChatGPT model — GPT-5.3 Instant. The update aims to make everyday conversations more natural, helpful, and consistent, while making responses more accurate and appropriate.

GPT-5.3 Instant is less likely to issue excessive refusals and lengthy warnings that previously hindered conversations. If a request allows for a direct and safe response, the model now more often gets straight to the point without unnecessary caveats, reducing the number of "dead-end" dialogues.

The update also improves handling of information from the internet: the model better matches data from web sources with its own knowledge and logic, explaining the context of news and events rather than simply retelling search results. This allows for faster identification of key points and more meaningful answers.

Additionally, GPT-5.3 Instant has adopted a more natural conversational style. Responses have become less intrusive and more focused, without unnecessary emotional assumptions or formulaic phrasing. OpenAI emphasizes that the goal of the update is to preserve the familiar ChatGPT experience while making it smoother and more consistent across different dialogues.

The new version also demonstrates higher accuracy and a reduced tendency toward hallucinations. Internal evaluations show a decrease in errors both when using web search and when answering based solely on the model's knowledge.

GPT-5.3 Instant is already available to all ChatGPT users, as well as to developers via the API. The previous version of the model will remain available to paying users in the Legacy Models section until June 2026.

Source: openai.com