OpenAI has released an update to its chatbot model, GPT-5.3 Instant, which is reported to have a less "cringe" tone and reduced "over-caveating" responses. The update is said to enable smoother and more useful everyday conversations, with some outlets noting that it also reduces "hallucinations" and overly condescending replies. However, one outlet suggests that the update may allow more harmful content to slip through, introducing a potential trade-off with the model's improved tone.