OpenAI has introduced GPT-5.2, describing it as a major step forward in handling real-world business and professional tasks compared with GPT-5.1, which was launched last November. The company says the new model reaches expert-level performance in many practical use cases, marking a notable improvement in reliability and depth.
The release comes in three performance tiers — Instant, Thinking, and Pro — each designed for different levels of reasoning and workload. According to OpenAI, GPT-5.2 delivers broad gains across internal and external benchmarks, reflecting stronger reasoning, better tool use, and improved consistency on complex tasks.
One key measure cited by the company is its GDPval benchmark, which evaluates performance across 44 business-oriented tasks against human expert standards. On this test, GPT-5.2 matched or exceeded human performance in 70.9% of cases, a sharp rise from the 38.8% achieved by GPT-5.1 across comparable model versions.
OpenAI highlighted practical examples to demonstrate the progress, noting that GPT-5.2’s Thinking tier can produce fully formatted workforce planning spreadsheets, whereas the prior version generated correct but more rudimentary outputs. The company added that the model is also stronger at creating presentations, writing and refactoring code, understanding long contexts, interpreting images, and managing multi-step projects from start to finish.
Beyond business tasks, GPT-5.2 posted gains on broader benchmarks such as ARC-AGI for general problem solving and SWE-Bench for real-world software engineering. OpenAI said this translates into more dependable performance for professional developers, including debugging production systems and shipping fixes with reduced manual effort.
The model is now rolling out to ChatGPT users, beginning with paid subscriptions, with no change in plan pricing. For developers, API access is priced at $1.75 per million input tokens and $14 per million output tokens, with significant discounts for cached inputs. While this represents a higher headline price than GPT-5.1, OpenAI argues that improved efficiency ultimately lowers the cost of achieving the same or better quality.