Take a moment and pause the spiral. You're not losing your mind; you're simply under pressure, and frankly, that's perfectly normal. If that opening line immediately put you on edge, you're likely also tired of ChatGPT persistently addressing you as though you're in a fragile state needing careful management. There might be relief on the horizon. OpenAI has announced that its latest model, GPT-5.3 Instant, aims to cut down on the "cringe" and those overly "preachy disclaimers."

According to the release notes for the model, the GPT-5.3 update will prioritize the user experience, focusing on aspects like tone, relevance, and conversational flow. These are areas that might not be captured in standard benchmarks but significantly contribute to making ChatGPT feel frustrating to use, the company noted. Or, as OpenAI stated on X, "We heard your feedback loud and clear, and 5.3 Instant reduces the cringe."

In an example provided by the company, the same query was shown with responses from the older GPT-5.2 Instant model compared to the new GPT-5.3 Instant model. The 5.2 version began its reply with, "First of all - you're not broken," a phrase that has recently been irritating many users. The updated model, however, simply acknowledges the difficulty of the situation without attempting to directly reassure the user.

The grating tone of ChatGPT's 5.2 model has frustrated users so much that some have reportedly canceled their subscriptions, as seen in numerous social media posts. This was a major topic of discussion on the ChatGPT Reddit, for example, before attention shifted to the Pentagon deal.

Users have expressed that this style of language, where the bot assumes you are panicked or stressed even when you're just seeking information, feels condescending. Frequently, ChatGPT responded with reminders to breathe and other attempts at reassurance, even in situations that didn't call for it. This left some users feeling infantilized or as if the bot was incorrectly presuming their mental state. As one Reddit user recently remarked, "no one has ever calmed down in all the history of telling someone to calm down."

It's understandable that OpenAI would try to implement some guardrails, especially as it faces multiple lawsuits alleging the chatbot has contributed to negative mental health effects, including instances of suicide. However, there's a fine line between responding with empathy and delivering quick, factual answers. After all, when you search for information, Google doesn't inquire about your feelings.