OpenAI has pushed a user-facing tweak to ChatGPT that aims to make conversations less awkward and more useful — a change that could matter to crypto teams and traders who rely on the chatbot for quick answers, debugging and research.
What changed
- On Tuesday OpenAI began rolling out GPT-5.3 Instant as the default ChatGPT model. The update doesn’t add major new capabilities; it focuses on day-to-day interaction quality: fewer overly cautious refusals, fewer preachy or repetitive disclaimers, and more direct answers.
- OpenAI summarized the goal crisply on X: “More accurate, less cringe.” A company spokesperson told Decrypt the release responds to widespread user feedback that prior models could be stiff or long-winded in normal conversations.
Accuracy and hallucinatory behavior
- Alongside tone changes, OpenAI reports measurable improvements in factual reliability. Internal evaluations show hallucination rates dropped by nearly 30% in some settings:
- On a higher-stakes evaluation, hallucinations fell 26.8% when the model used the web and 19.7% when it relied only on internal knowledge, versus earlier models.
- In a user-feedback evaluation, hallucinations declined 22.5% with web access and 9.6% without it.
Design choices and trade-offs
- OpenAI says GPT-5.3 Instant reduces cases where the model needlessly declines questions or leads with safety preambles. The company contrasted GPT-5.2’s habit of opening with lengthy safety caveats with GPT-5.3’s approach of getting straight into the answer.
- OpenAI also highlighted improved writing skill (it compared poetry generated by 5.2 and 5.3), though it didn’t define what it meant by “cringe.”
Rollout and availability
- GPT-5.3 Instant is now the default ChatGPT model. GPT-5.2 Instant remains available under legacy options for paid subscribers during a transition window that ends in early June.
Community reaction and context
- Reaction on social media was mixed. Some users praised the shift toward more direct answers with fewer boilerplate caveats. Others criticized OpenAI’s broader decisions — notably the company’s contract with the U.S. Department of Defense — arguing that the “real cringe” is the partnership, a point raised after competitors like Anthropic opted out over safety concerns.
- Some long-time users also said GPT-5.3 lacks the intimacy they associated with the now-deprecated GPT-4o and asked for that model’s return.
- The rollout follows last summer’s controversy when OpenAI replaced GPT-4o with GPT-5, a move that drew strong pushback and briefly prompted OpenAI to restore GPT-4o for paid users. OpenAI CEO Sam Altman later said the episode was a “lesson in upgrading a product used by hundreds of millions of people.”
Why crypto readers should care
- Many crypto developers, security auditors and traders use ChatGPT for code help, research and rapid prototyping. A model that refuses less often, gives more direct answers and hallucinates less could speed workflows — but the community response shows trust and safety concerns remain central to adoption.
Note: this story was updated after publication to include comments from OpenAI.
Read more AI-generated news on: undefined/news
