OpenAI launches GPT-5.2

OpenAI has launched GPT-5.2, their latest flagship model made for professional, agentic work. Its improvements from the previous model include stronger general intelligence, better context handling, and upgraded tools and vision capabilities.

GPT-5.2 Thinking now outperforms industry professionals on 70.9 percent of GDPval benchmark tasks across 44 occupations. It’s also capable of delivering outputs at over 11x the speed and under 1 percent of the cost of human experts.

It also posts new highs on coding benchmarks such as SWE-Bench Pro and SWE-bench Verified. This makes it more capable of debugging, refactoring, and shipping production-ready code end-to-end.

The model leads in long-context and tool-calling tests like MRCRv2 and Tau2-bench. This translates into more reliable multi-step workflows spanning large documents along with multiple tools.

Vision has also been upgraded, with roughly half the error rate of previous models on chart reasoning and UI understanding tasks. GPT-5.2 is rolling out via ChatGPT as Instant, Thinking, and Pro.