OpenAI Releases GPT-5.3 Instant to Cut Down on Preachy, Condescending Responses
The update to ChatGPT's default model targets tone and conversational flow, reducing unsolicited emotional reassurances that frustrated users for months.
OpenAI on Monday released GPT-5.3 Instant, an update to ChatGPT's most widely used model focused on improving tone, relevance, and conversational flow. The company described the release as directly addressing months of user complaints about ChatGPT offering unsolicited emotional reassurances and moralizing caveats in response to straightforward questions.
"We heard your feedback loud and clear, and 5.3 Instant reduces the cringe," OpenAI said in a post on X. The company's release notes showed the contrast starkly: where GPT-5.2 Instant responded to a user's work stress by opening with "First of all — you're not broken," the updated model acknowledged the situation more neutrally without attempting to emotionally manage the user.
The earlier model had produced responses such as "Take a breath, stop spiraling. You're not crazy, you're just stressed.
And honestly, that's okay" — a tone that had been widely mocked online and cited as a reason for user frustration.
The update focused on areas that do not show up in standard benchmarks — tone, appropriateness of disclaimer density, and conversational naturalness — rather than on raw capability metrics. OpenAI said the model would reduce unnecessary dead ends, excessive caveats, and overly declarative phrasing that can interrupt the flow of conversation.
The release came at an already-turbulent moment for OpenAI, with the company managing backlash from its Pentagon deal announced the previous week. Some observers noted the timing as coincidental but pointed to broader pressure on OpenAI to refine its product experience as competition from Anthropic and Google intensified.
GPT-5.3 Instant was made available to all ChatGPT users, including those on the free tier.
Read the original reporting at TechCrunch.
