OpenAI debuts GPT-5.2, aiming to transform professional workflows

WEB DESK: OpenAI has released GPT-5.2, calling it its most advanced and professionally capable model yet — one built to handle complex knowledge work, long-running autonomous tasks, and high-level workflows that usually require expert teams. The company says the model marks a turning point in AI’s ability to perform real professional labour rather than simply assist with it.

Performance That Rivals Top Professionals

GPT-5.2’s “Thinking” variant delivers a dramatic jump in performance on GDPval, a benchmark covering 44 professional occupations. According to OpenAI, the model now matches or surpasses leading human professionals in more than 70 per cent of tasks, nearly doubling GPT-5’s 38.8 per cent. These tasks include building financial models, drafting technical documents, designing manufacturing diagrams and generating polished presentations — all at speeds far beyond human capability. One evaluator reviewing its output remarked that the work “looked like it was produced by a full professional team.”

Major Gains in Engineering and Technical Work

The model also records a new high on SWE-Bench Pro, the toughest benchmark for real-world software engineering. GPT-5.2 can more reliably debug deployed code, refactor large projects and implement intricate feature requests. Tech partners such as Windsurf and JetBrains describe this release as the “biggest leap yet” in AI coding agents, especially noting improvements for front-end development and even complex 3D user interfaces.

More Reliable, More Context-Aware

OpenAI says GPT-5.2 is significantly more stable than its predecessors, with a 30 per cent reduction in user-visible errors. It also achieves near-perfect performance on long-context evaluations, accurately processing and synthesizing information scattered across documents of up to 256,000 tokens. This capability makes it a powerful tool for analysing long contracts, research papers, or multi-file technical repositories in a single session.

Sharper Vision for Real-World Tasks

The model’s visual reasoning has also been upgraded. Error rates on tasks involving charts, dashboards, UI screenshots and technical diagrams have been reduced by roughly half. This allows GPT-5.2 to deliver more accurate insights across finance, design, product development, customer support and other fields that rely heavily on visual data.

Rolling Out Now Across API and ChatGPT

GPT-5.2 is available immediately through the OpenAI API. ChatGPT’s paid users will begin seeing the new model appear across the Instant, Thinking and Pro variants, each equipped with upgraded tools for producing advanced spreadsheets and presentation-ready documents directly from prompts.

A Shift Toward Full Workflow Automation

OpenAI is framing GPT-5.2 as more than a routine upgrade, instead, it is a system engineered to run entire professional workflows from start to finish. With stronger reasoning, deeper context handling and more reliable execution, GPT-5.2 pushes AI closer to functioning as a true digital collaborator capable of delivering end-to-end work across multiple domains.