OpenAI rolls out GPT-5.2, its most advanced model yet

OpenAI has introduced GPT-5.2, its most advanced AI model yet, and says the company expects to move out of its internal “code red” mode by January. The new model is designed to handle professional tasks more effectively, including creating spreadsheets, building presentations, interpreting images, writing code and working with long documents.

GPT-5.2 is now available inside ChatGPT and through OpenAI’s API. The launch comes only weeks after GPT-5.1, and follows similar upgrades from competitors Google and Anthropic, pressure that pushed OpenAI to temporarily shift its focus to improving ChatGPT.

The “code red” designation was meant to rally the company around upgrading its flagship product, according to Fidji Simo, OpenAI’s CEO of applications. She said the company concentrated more resources on ChatGPT but stressed that GPT-5.2’s release timeline wasn’t tied specifically to Google’s launch.

Speaking to CNBC, CEO Sam Altman added that Google’s new Gemini 3 model didn’t impact OpenAI’s metrics as much as expected, and said he believes the company will return to normal operations early next year.

OpenAI is offering GPT-5.2 in three versions: Instant, Thinking and Pro. Instant focuses on speed, Thinking is built for structured tasks like coding or planning, and Pro is meant to deliver the most accurate results for complex questions.

The company says the model leads industry benchmarks, outperforming competitors on tests for coding, scientific reasoning and real-world problem solving. However, Anthropic’s latest Opus 4.5 model still scores higher on one widely cited coding benchmark, though OpenAI argues that the test is less reliable than others.