OpenAI Answers Google With GPT-5.2, Weeks After an Internal 'Code Red'

The new model is pitched at office work — spreadsheets, presentations, long documents, multi-step projects — and arrives after reports that Google's Gemini 3 sent OpenAI's leadership into emergency mode.

OpenAI released GPT-5.2 today, three weeks after Google put out a model that beat it. The company is positioning this one squarely at professional work: it is better, OpenAI says, at building spreadsheets and presentations, reading long documents, interpreting images, using other software, and carrying out projects that take many steps. It comes in three settings — a fast one, a thinking one, and a heavier "Pro" version for hard problems.

The context is a memo. After Gemini 3 launched in November and took the lead on the public benchmarks, Sam Altman reportedly declared an internal "code red" at OpenAI, reorganizing teams around the competitive threat. Google, which spent 2023 and 2024 being described as the company that invented this technology and then lost the lead, is now the one setting the pace often enough that its rival is holding emergency meetings about it.

For most people the practical difference between these models is small and getting smaller. All three leading labs now ship systems that are roughly as capable, a few weeks apart, at steadily falling prices. The competition is fierce and the products are converging — which is usually what happens to a technology on its way to becoming ordinary.

What is not ordinary is the pace of the exchange. Google in November, Anthropic six days later, OpenAI three weeks after that. A field where the state of the art changes hands monthly is a field where nobody — including the people building it — has a stable picture of what is possible.