'Extra Correct, Much less Cringe': OpenAI Rolls Out GPT-5.3 Prompt in ChatGPT – Decrypt

OpenAI on Tuesday introduced the rollout of GPT-5.3 Prompt, an replace to ChatGPT’s default mannequin geared toward making conversations really feel much less awkward and preachy, and “extra straight” useful.

In a product put up, OpenAI stated the brand new model reduces overly cautious refusals, trims pointless disclaimers, and delivers extra correct solutions. The adjustments mirror consumer complaints that earlier variations may sound stiff or overbearing in on a regular basis interactions.

“Extra correct, much less cringe,” OpenAI wrote on X. “We heard your suggestions loud and clear.”

Somewhat than introducing new capabilities, the replace targets routine interactions. OpenAI stated earlier variations typically declined questions they might safely reply, or interrupted responses with prolonged explanations about security limits.

“GPT‑5.2 Prompt ultimately solutions the query, however in an try to clarify its security boundaries, leads with a prolonged preamble about what it can’t assist with,” OpenAI wrote. “GPT‑5.3 Prompt, however, will get proper into the response.”

OpenAI reported enhancements in factual reliability alongside the tone adjustments, claiming that inside evaluations confirmed hallucination charges dropped by practically 30%.

“On the higher-stakes analysis, GPT‑5.3 Prompt reduces hallucination charges by 26.8% when utilizing the online and 19.7% when relying solely on its inside information, in comparison with prior fashions,” OpenAI stated. “On the user-feedback analysis, hallucinations lower by 22.5% with net use and 9.6% with out net entry.”

OpenAI didn’t clarify what it defines as “cringe,” however famous that the brand new mannequin consists of stronger writing talents, evaluating GPT-5.2 and 5.3’s means to write down poetry.

“5.4 earlier than you suppose,” the corporate stated in a separate put up, which drew swift mockery from customers on X, suggesting that the tease was on account of current backlash towards the agency for its take care of the Pentagon.

GPT-5.3 Prompt replaces the default ChatGPT mannequin beginning at present, the corporate stated, whereas GPT-5.2 Prompt stays accessible underneath legacy choices for paid subscribers throughout a transition interval ending in early June.

The replace drew blended reactions on social media. Some customers praised the give attention to extra direct responses with out pointless disclaimers, whereas others argued the actual “cringe” at play was agreeing to a contract with the U.S. Division of Protection when rival Anthropic declined on account of security considerations.

Others complained that GPT-5.3 would by no means match the sense of intimacy many related to the now-depreciated GPT-4o, and known as for the favored mannequin’s return.

Final summer season, OpenAI confronted a surge in backlash after the corporate abruptly changed the favored GPT-4o with GPT-5, prompting complaints that the brand new mannequin felt colder and fewer supportive. Customers flooded boards with criticism, and a few threatened to cancel subscriptions, main OpenAI to revive GPT-4o for paid customers.

In January, OpenAI introduced that GPT-4o and its variants could be formally retired as of February 13.

“I believe we’ve discovered a lesson about what it means to improve a product for lots of of thousands and thousands of individuals in someday,” OpenAI CEO Sam Altman stated on the time, calling the reversal a wake-up name and “a lesson in upgrading a product utilized by lots of of thousands and thousands of individuals” without delay.

OpenAI didn’t instantly reply to a request for remark by Decrypt.