OpenAI has announced an update to its GPT-5.3 Instant model for ChatGPT, designed to reduce responses that many users have found irritating. The company aims to improve the overall experience by eliminating what it calls "cringe" responses.

Context

Large language models (LLMs) like those underlying ChatGPT are constantly evolving. Companies are constantly working to refine responses, balancing utility, safety, and accuracy. This process includes correcting unwanted behaviors that emerge during user usage. The challenge is to calibrate the model to provide relevant and useful responses, while avoiding overly cautious or patronizing reactions.

For those evaluating on-premise deployments, there are trade-offs to consider. AI-RADAR offers analytical frameworks on /llm-onpremise to evaluate these aspects.