OpenAI has released GPT‑5.3 Instant, a significant update to ChatGPT’s most-used model, designed to make daily interactions more fluid, helpful, and contextually aware. This version addresses key areas highlighted by user feedback: tone, relevance, and conversational flow, reducing unnecessary caveats, disclaimers, and dead ends that could disrupt communication.
GPT‑5.3 Instant improves judgment in sensitive topics by providing direct answers when appropriate, minimizing overly cautious or moralizing preambles. When accessing information online, the model now balances search results with its internal knowledge, prioritizing relevant information upfront and avoiding long, loosely connected lists. The update also refines conversational style, cutting overbearing or awkward phrasing, while maintaining consistency across interactions. Users can still adjust the model’s tone for warmth or enthusiasm. Beyond factual responses, GPT‑5.3 Instant enhances creative and expressive writing, making it a more versatile partner for both practical tasks and imaginative projects.
Accuracy has increased, with hallucination rates dropping by 26.8% in high-stakes domains like medicine, law, and finance when using the web, and by 19.7% when relying solely on the model’s internal knowledge. In user-feedback evaluations, hallucinations decreased by 22.5% with web use and by 9.6% without it.
GPT‑5.3 Instant is available today for ChatGPT users and developers via the API as gpt-5.3-chat-latest, with updates to Thinking and Pro planned soon. GPT‑5.2 Instant will remain accessible for paid users until June 3, 2026, after which it will retire.