OpenAI is facing massive public backlash after it signed a deal with the Pentagon over the use of its technology in classified military operations. And one of the first customer-facing actions the company took was... releasing a new small GPT model that’s supposed to be “less cringe” compared to its predecessors.
OpenAI announced the release of GPT-5.3 Instant today, an update for one of the most popular GPT models, GPT-5.2 Instant. The highlight of this update is that the new model is now more natural, less dramatic in its responses, and reduces the number of unnecessary refusals to users’ questions.
GPT‑5.3 Instant also has better reasoning over web searches, so it should provide you with more accurate summaries of search results when you ask it to browse the web. OpenAI also claims that the new model hallucinates less and delivers more factual responses than previous models.
GPT‑5.2 Instant’s tone could sometimes feel "cringe," coming across as overbearing or making unwarranted assumptions about user intent or emotions.
This update has a more focused yet natural conversational style, cutting back on unnecessary proclamations and phrases like "Stop. Take a breath." We’re also working to keep ChatGPT’s personality more consistent across conversations and updates.
GPT‑5.3 Instant is now available to all users, so you don’t have to be a premium subscriber to try it.
Truth be told, cringe is the signature trait of AI, so any update that decreases the level of drama in a model is welcome. However, just because the GPT-5.3 Instant is “less cringe” on paper than the previous one, doesn’t mean it won"t still sound like a robot trying too hard to be smart. So, don’t expect a literary genius; it’s still just an AI model, and a small one at that.
You can learn more details about the GPT-5.3 Instant model in OpenAI’s blog post.