OpenAI Launches GPT-5.2 With Big Leaps in Reasoning, Coding & Real-World Tasks

OpenAI today announced the release of GPT-5.2, the newest model in its GPT-5 family, designed to deliver significant improvements in intelligence, reasoning, coding, and long-context understanding for real-world applications.

The update is aimed at professional knowledge work, complex multi-step tasks, and enhanced productivity across industries. According to OpenAI, GPT-5.2 outperforms both its predecessors and many human experts on a range of benchmarks.

In tests such as GDPval, which measures performance on practical work tasks across 44 professions, GPT-5.2 achieved a new state-of-the-art score, completing work like presentations, spreadsheets and analytical reports much faster and more accurately than previous models.

The model is rolling out today in ChatGPT for paid subscribers as GPT-5.2 Instant, GPT-5.2 Thinking and GPT-5.2 Pro, each optimized for different use cases. GPT-5.2 Instant excels in everyday interactions, while the Thinking and Pro variants offer deeper reasoning, complex planning and improved tool integration. Initial availability is focused on paid tiers, with legacy models like GPT-5.1 remaining accessible for a limited transition period.

OpenAI highlighted that GPT-5.2’s enhancements also include stronger code generation, better structured outputs, fewer errors and expanded visual and multi-modal understanding. Developers can access the updated models immediately via OpenAI’s API, including responses and chat endpoints.

In a statement, users early testing GPT-5.2 noted its ability to tackle complex engineering tasks and long document analysis much more reliably, reflecting OpenAI’s goal of making advanced AI agents more useful and dependable for enterprise workflows.

The rollout underscores OpenAI’s focus on practical AI adoption, equipping tools that aid researchers, engineers and professionals tackling demanding, real-world challenges.