OpenAI has launched an updated version of its most frequently used ChatGPT model, GPT-5.3 Instant, aimed at making daily conversations "more stable, helpful, and fluid."

This update reduces instances of unnecessary refusals to answer and tones down "overly defensive or preachy preambles" that previously preceded responses.

It enhances the quality of answers when information is sourced from the web.

The model more effectively balances information retrieved from online searches with its own internal knowledge and reasoning capabilities.

It demonstrates a stronger ability to discern the underlying intent of questions and prioritizes presenting the most critical information first.

Compared to previous versions, it provides responses that are more factually grounded and exhibits a reduction in hallucinatory content across a wide range of topics.

The update will be available to all ChatGPT users and API developers starting March 3rd.