The San Francisco-based AI superstar touted GPT-5.2 Pro and GPT-5.2 Thinking as its best models yet for handling math or science work.

"Strong mathematical reasoning is a foundation for reliability in scientific and technical work," OpenAI said in a blog post.

"These capabilities are also closely tied to progress toward general intelligence."

Artificial general intelligence has become a holy grail of sorts in the tech world, seen as a threshold where machines think the way people do or even better.

The release comes on the heels of OpenAI chief executive Sam Altman urging his team to strive to keep up with Google, the search engine juggernaut that has been relentlessly innovating in AI.

While Google can tap into its massive online ad revenue to invest in AI, OpenAI has been committing tens of billions of dollars to computing infrastructure while having yet to turn a profit.

"We are confident we can continue to drive the revenue growth to meet" the investments in computing power, Altman said Thursday in a CNBC interview.

Without the infrastructure investments, "of course, we can't drive the revenue growth, but we see way more reasons to be optimistic than reasons to be pessimistic."

OpenAI chief of applications Fidji Simo told reporters during a briefing about the new models that she expects a ChatGPT "adult mode" to debut early next year, noting that the company wants to improve detection of user age before making it available.

Altman earlier this year announced plans to ease restrictions to allow adult users to engage in erotic conversations with ChatGPT.

OpenAI also faces a series of lawsuits from families accusing the startup of allowing teenagers to have dangerous interactions with its AI chatbots that in some cases led to suicide.

Simo confirmed that a "red alert" about Google sprinting ahead had been issued at OpenAI, but refuted the notion it has sped up the release of new GPT models.

Google last month debuted its latest Gemini AI model, capping a dramatic turnaround since it was caught off guard by ChatGPT's launch three years ago and mocked for early blunders in its chase of OpenAI.