OpenAI Launches GPT-5.2: The most powerful AI model yet arrives in ChatGPT and API

The model brings improvements in creating tables and presentations, processing and interpreting images, programming and managing especially long and complex instructions, segments in which OpenAI wants to maintain an advantage over the competition.

The race is heating up

The presentation will come just a few weeks after the release of GPT-5.1. Although the pace of releases seems accelerated, the reason is competition. OpenAI has found itself in a fierce race in recent months with others aggressively pushing their own models, including Anthropic and Google. Their updates prompted OpenAI to, as they call it, a ‘red alarm’ and divert part of their internal resources to strengthening ChatGPT.

Fidji Simo, head of OpenAI’s applications, said that the internal directive was clear – to focus on one priority.

– That was a way to clarify what must have priority and what can wait – she said, adding that this focus did not dictate the exact date of the new model’s launch.

Sam Altman, CEO of OpenAI, revealed that the launch of Google’s Gemini 3 ended up having less of an impact on their users than they expected, and that the company should come out of “red alert” at the beginning of September.

Three versions for three types of users

GPT-5.2 comes in three variants:

OpenAI claims their new model tops key tests, including SWE-Bench Pro for autonomous coding and GPQA Diamond, a test of advanced scientific reasoning. In their own GDPval testing, which they previously presented to the public, they said that GPT-5.2 was equal to or better than the leading experts in more than 70% of clearly defined tasks.

Months of work and tough competition

Simo emphasizes that GPT-5.2 has been in the works for months, regardless of čest announcements of new models. Integration, he claims, was not fast, butć thorough.

But the competition is not sleeping, the newly presented Anthropic’s Opus 4.5 surpassed GPT-5.2 on the SWE-Bench Verified test, but OpenAI says that the test is narrowly specialized and less reliable for evaluating real applications than the SWE-Bench Pro benchmark.

OpenAI, which started as a research laboratory ten years ago, is today a global AI force thanks to the explosion of popularity of ChatGPT in 2022. The company states that their chatbot is used by more than 800 million people every week, making them one of the most influential players in the gaming industry.