ChatGPT's awkward, overly dramatic responses are getting a major overhaul with GPT-5.3 Instant, an update designed specifically to address user complaints about the AI assistant's "cringe" conversational style.
OpenAI rolled out the new model to all ChatGPT users this week, promising more natural interactions that cut back on unnecessary proclamations and jarring phrasing. The company says GPT-5.3 Instant "delivers more accurate answers, richer and better-contextualized results when searching the web, and reduces unnecessary dead ends, caveats, and overly declarative phrasing that can interrupt the flow of conversation."
Previous versions sometimes felt overbearing or made unwarranted assumptions about user intent or emotions, according to OpenAI's release notes. The new model tones down defensive or moralizing preambles before answering questions and eliminates phrases like "Stop. Take a breath." that became meme material.
Technical improvements include better integration of web-sourced information with internal knowledge bases.
In higher-risk domains like medicine, law and finance, hallucination rates decreased by 26.8% when using web browsing and 19.7% when relying solely on internal knowledge, according to company testing. User-reported factual errors dropped by 22.5% with web access enabled.
The update arrives as ChatGPT faces increased competition from Anthropic's Claude, which has seen daily sign-ups hit record highs and free users jump by more than 60% since January following controversies around OpenAI's Pentagon agreement earlier this year.
GPT-5.3 Instant is already available through the API under gpt-5.3-chat-latest for developers building on OpenAI's platform. Existing GPT 5.2 will remain accessible as a legacy option for paid users through June 3 before being retired.
