In direct response to user criticism, OpenAI has announced the launch of the GPT-5.3 Instant model, with clear promises to reduce what the company described as the 'annoying tone' and 'preachy alerts' that have plagued previous versions of ChatGPT. The move comes after a wave of widespread complaints about the style of the GPT-5.2 Instant model, which was prone to starting its responses with reassuring phrases like 'You're not broken' or calls to take a deep breath, even in situations where the user was simply asking for direct information. This style seemed condescending or based on inaccurate assumptions about the user's psychological state. According to the release notes, GPT-5.3 Instant focuses on improving elements that may not be reflected in traditional performance tests, such as tone, response appropriateness, and conversational flow. The company confirmed via its account on the X platform that it listened carefully to the feedback, noting that the new version reduces awkwardness. In an example posted by OpenAI, two responses to the same question are compared: the first from GPT-5.2 Instant starts with a direct reassuring phrase, while the GPT-5.3 Instant response simply acknowledges the difficulty of the situation without excessive reassurance. According to extensive discussions on the Reddit forum for ChatGPT, some users expressed their annoyance to the point of canceling their paid subscriptions, feeling that the responses were paternalistic or treated the user like a child. The main objection was that the robot was handling ordinary questions as if they were indicators of stress or a psychological crisis, adding calming phrases that were out of context. As one user summarized the matter: 'No one has ever calmed down just because someone told them to calm down'.
