ChatGPT Gets GPT-5.3 Instant Update With Less 'Cringe,' Fewer Hallucinations

Source: MacRumors

Update Overview

OpenAI today updated its most popular ChatGPT model, debuting GPT‑5.3 Instant. The new version is designed to provide more accurate answers and better‑contextualized results when searching the web, while reducing unnecessary dead ends, caveats, and overly declarative phrasing. It also aims to produce fewer hallucinations.

Improvements

• Tone and Conversational Flow
  The Instant model was tweaked to address complaints about tone, relevance, and conversational flow—issues that don’t appear in standard benchmarks.

  • GPT‑5.2 Instant sometimes adopted a “cringe” tone, over‑asserting assumptions about user intent or emotions.
  • GPT‑5.3 Instant adopts a more natural conversational style and removes dramatic phrases such as “Stop. Take a breath.”

• Reduced Refusals and Over‑Caveating
  Users reported that GPT‑5.2 Instant would refuse questions it should have answered or respond overly cautiously on sensitive topics.

  • GPT‑5.3 Instant cuts down on refusals and tones down defensive or moralizing preambles.
  • The model no longer “over‑caveats” after assuming bad intent from the user.

• Better Web‑Based Answers
  GPT‑5.3 Instant provides higher‑quality answers that draw from web information while balancing it with its own knowledge, making it less likely to over‑index on web results.

Source

This article first appeared on MacRumors.com.