Daijiworld Media Network - New Delhi
New Delhi, Dec 12: Days after internally declaring a ‘code red’ over Google’s growing edge in artificial intelligence, OpenAI has hit back with the launch of GPT-5.2 — its most advanced frontier model designed for professional work and long-running AI agents.
The Sam Altman-led company says GPT-5.2 marks a major leap in enterprise-grade AI, offering faster performance, stronger reasoning and far greater reliability across real-world tasks than previous models.
What is GPT-5.2?
GPT-5.2 is a next-generation AI model engineered for workplace productivity. It significantly improves tasks such as summarising long documents, writing and debugging code, building spreadsheets and presentations, analysing images, managing data, and completing multi-step projects. OpenAI says companies using ChatGPT Enterprise have already saved up to an hour a day — savings expected to rise sharply with GPT-5.2.
The model is better at reading extremely long documents, maintaining context across hundreds of thousands of tokens, solving complicated problems without losing track, and calling external tools such as databases or company software.
GPT-5.2 comes in three versions on ChatGPT — Instant for fast everyday use, Thinking for deeper structured reasoning, and Pro for the highest-quality responses on complex tasks. It is rolling out across paid plans and is already available in the API as gpt-5.2, gpt-5.2-chat-latest and gpt-5.2-pro.
Performance breakthroughs
OpenAI revealed that GPT-5.2 excelled in GDPval, a test spanning 44 professional domains. The ‘Thinking’ version matched or exceeded human experts on 70.9% of tasks — nearly double GPT-5’s performance.
In coding, GPT-5.2 set a new record on the SWE-Bench Pro benchmark, showing improved debugging, feature implementation, code review and end-to-end engineering capabilities. Developers also found it far better at front-end tasks, generating complex visuals and 3D interfaces with fewer errors.
Beyond text and coding, GPT-5.2 shows stronger vision skills, interpreting charts, dashboards, UI screenshots and even poor-quality images with greater accuracy.
Multi-step task mastery
The model scores 98.7% in Tau2 benchmarks for telecom customer support, showing it can handle multi-step workflows with planning and tool-use without dropping tasks midway. It reportedly managed complex operations such as travel rebooking, locating luggage, arranging hotels and processing medical-seating requests — all in one continuous thread.
For professionals handling legal documents, contracts, research papers or multi-file projects, GPT-5.2 is designed to process huge datasets without the need to manually break them down.
The model also boasts major gains in scientific and mathematical reasoning, achieving over 92% on graduate-level science questions and setting a new record on advanced math problems. Researchers have even used it to propose proofs in statistical learning theory, later verified by experts.
What it means for OpenAI
The release comes amid fierce competition from Google’s Gemini 3 and Anthropic’s Claude Opus 4, both strong performers in recent benchmarks. After Gemini 3’s rise, Altman reportedly declared a company-wide ‘code red’, urging employees to focus on chatbot quality while delaying non-essential plans, including ads.
With GPT-5.2, OpenAI aims to deliver more economic value to users, particularly in areas requiring spreadsheets, presentations, planning and multi-step project management — all crucial in the enterprise market it seeks to dominate.