OpenAI launches GPT-5.2, its most powerful model yet with big upgrades in coding, image understanding, and long-context reasoning

OpenAI on Thursday introduced GPT-5.2, calling it its most capable model yet for day-to-day professional work. The release comes only weeks after GPT-5.1, marking the company’s fastest model-to-model shift to date as competition in AI development intensifies.

GPT-5.2 brings upgrades across some of the tasks people rely on most. OpenAI says it handles spreadsheets more accurately, builds cleaner presentations, reads images with better precision, writes stronger code, and works with long context far more reliably than earlier versions. The model will begin rolling out on Thursday in ChatGPT and via the API.

“We are introducing GPT‑5.2, the most capable model series yet for professional knowledge work,” OpenAI said in a post on Thursday.

According to OpenAI, GPT-5.2 Thinking is the strongest model we’ve released for real-world professional work. On GDPval — an evaluation built around well-defined knowledge-work tasks spanning 44 occupations — the model sets a new state of the art and becomes our first system to match or exceed human experts consistently.

“GPT‑5.2 Thinking beats or ties top industry professionals on 70.9% of comparisons on GDPval knowledge work tasks, according to expert human judges. These tasks include making presentations, spreadsheets, and other artifacts. GPT‑5.2,” OpenAI said.

According to expert judges, GPT-5.2 Thinking beats or ties top industry professionals in 70.9% of head-to-head comparisons. The evaluation covers practical assignments such as building presentations, creating spreadsheets, and producing other work artifacts.

GPT-5.2 Thinking also generated outputs for GDPval tasks more than 11× faster and at under 1% of the cost of expert professionals. These figures are based on historical benchmarks, and actual ChatGPT speed may vary, but they illustrate how pairing GPT-5.2 Thinking with human oversight can meaningfully support professional workflows.

GPT-5.2 Thinking evals

The timing raised questions because Anthropic and Google pushed out new models last month, prompting OpenAI leaders to declare a “code red” to sharpen their focus on ChatGPT. Fidji Simo, CEO of applications at OpenAI, told reporters the shift helped align teams but didn’t dictate the release date, CNBC reported.

“We announced this code red to really signal to the company that we want to martial resources in one particular area, and that’s a way to really define priorities and define things that can be deprioritized,” she said. “We have had an increase in resources focused on ChatGPT in general. I would say that helps with the release of this model, but that’s not the reason it’s coming out this week in particular.”

GPT-5.2 arrives in three modes: Instant, Thinking, and Pro. Instant focuses on speed for writing and quick information requests. Thinking delves deeper into structured tasks such as planning and coding. Pro is designed for users who need the highest accuracy for more challenging prompts and longer work sessions.

Simo said the team has been building GPT-5.2 for months, long before the “code red” made headlines inside the company. “This has been in the works for many, many months,” she said. “While we are proud that we are able to have a cadence of releasing models fast, this particular integration has been in the works for a while.”

The release lands at a moment when model performance has become a weekly storyline across the industry. With GPT-5.2, OpenAI is betting that faster iteration and a clear focus on ChatGPT will help keep pace with rivals—while giving everyday users a model that feels more reliable for real work.