OpenAI has released GPT-5.2, an upgrade the company says makes the model much better at real business tasks than GPT-5.1 from November. The new version comes in three levels: Instant, Thinking, and Pro.
On OpenAI’s GDPval test, which measures how well the model handles 44 business tasks at expert standards, GPT-5.2 matched or beat humans in 70.9% of the cases. By contrast, GPT-5.1 scored 38.8% across the Instant, Thinking, and Pro tiers.
OpenAI gave a clear example of the difference. GPT-5.2 Thinking can fully format a workforce planning spreadsheet. On GPT-5.1 the same spreadsheet was correct but basic and lacked formatting.
“We designed GPT‑5.2 to unlock even more economic value for people; it’s better at creating spreadsheets, building presentations, writing code, perceiving images, understanding long contexts, using tools, and handling complex, multi-step projects,” said OpenAI.
The company also reported gains on other tests such as ARC-AGI-1, ARC-AGI-2, SWE-Bench Pro, and SWE-Bench Verified. OpenAI says these gains make the model more reliable for common work tasks like debugging production code, adding new features, refactoring large codebases, and shipping fixes with less human work.
GPT-5.2 is starting to reach ChatGPT users, beginning with paid plans. Subscription prices have not changed. For API use, OpenAI set the rate at $1.75 per one million input tokens and $14 per one million output tokens, with a 90% discount on cached inputs. The company says the new model is more token efficient, so users can get a given level of quality at lower overall cost.
“the cost of attaining a given level of quality ended up less expensive due to GPT‑5.2’s greater token efficiency.”
Leave a comment