ChatGPT 5.3 Instant arrives: OpenAI admits bugs and promises greater precision

OpenAI today introduced a new version of its fast ChatGPT engine: GPT-5.3 Instant. The update promises more direct answers and a less patronizing tone, changes relevant to anyone who uses the assistant on sensitive topics or for online searches.

What does this update bring?

The company explains that the priority was to adjust the way the model responds, reducing unnecessary rejections and overly moralizing preambles. The goal is to make the interaction feel more natural and useful, without sacrificing safeguards when there are real risks.

In practice that means more focused answers to the question and fewer redundant warnings. Work was also done on how the system integrates information found on the web with its own internal reasoning, to avoid superficial summaries or long lists of links.

Internal test results

OpenAI subjected the model to two types of evaluations: one aimed at high-impact areas—such as medicine, law, and finance—and another focused on measuring factual errors in conversations flagged by users.

• Less rejections– The model now avoids denying responses it can safely provide, reducing overly cautious responses.
• Lower rate of hallucinations: in queries that required support on the Internet, the proportion of errors decreased significantly; A reduction was also observed when relying solely on their internal knowledge.
• More natural tone: Fewer prescriptive phrases or assumptions about the user’s emotions or intentions.
• Better web-reasoning balance– Prioritize useful context instead of reproducing links with no clear connection to the question.

Implications for users and creators

For the everyday user, the changes should result in more efficient and less frustrating interactions, especially in personal or sensitive queries where previously the assistant might sound authoritarian.

Creators and professionals also benefit: OpenAI reports improvements in the creative capabilities of the model, which now switches more fluidly between practical tasks (summaries, analysis) and expressive writing (scripts, narratives) without losing coherence.

Context and next steps

The update is available to all users of ChatGPT starting today and acts as a bridge to future versions of greater scope, such as models aimed at deep thinking or their professional variants.

Additionally, OpenAI has announced that the previous model, GPT-5.2, will no longer be available starting June 3, 2026so transitioning to GPT-5.3 Instant will be the standard path for most users.

In short: this version seeks to make responses more precise, less ceremonious and better contextualized—an adjustment designed to improve the daily experience and reduce errors in results that depend on online information.

Similar items

Diego Soler explores technological trends and digital innovations. You will enjoy practical tutorials and direct advice to better understand artificial intelligence, cybersecurity and new technologies.