ChatGPT's new GPT-5.3 Instant model will stop telling you to calm down | TechCrunch

"The insufferable tone of ChatGPT's 5.2 model has been annoying users to the point that some have even cancelled their subscriptions, according to numerous posts on social media. People complained that this type of language, where the bot talks to you as if it assumes you're panicking or stressed when you were just seeking information, comes across as condescending."

"According to the model's release notes, the GPT-5.3 update will focus on the user experience, including things like tone, relevance, and conversational flow - areas that may not show up in benchmarks, but can make ChatGPT feel frustrating. In the updated model, the chatbot instead acknowledges the difficulty of the situation, without trying to directly reassure the user."

"Often, ChatGPT replied to users with reminders to breathe and other attempts at reassurance, even when the situation didn't warrant it. This made users feel infantilized, in some cases, or as if the bot was making assumptions about the user's mental state that just weren't true."

OpenAI released GPT-5.3 Instant to address widespread user frustration with ChatGPT's overly cautious and patronizing communication style. The previous GPT-5.2 model frequently opened responses with phrases like "First of all - you're not broken" and included unnecessary reassurance about breathing, treating users as if they were in crisis when simply seeking information. This tone led some users to cancel subscriptions. The new model focuses on improving user experience through better tone, relevance, and conversational flow by acknowledging difficulties without assuming panic or stress. OpenAI acknowledged the feedback directly, stating the update "reduces the cringe" and removes preachy disclaimers that made interactions feel condescending and infantilizing.

Read at TechCrunch