OpenAI Unveils GPT-5.2 With Major Boosts in Coding, Reasoning and Workplace AI Performance

OpenAI has officially rolled out GPT-5.2, its newest frontier model designed to deliver significantly stronger performance in coding, reasoning, long-context reading and complex professional tasks. The launch comes days after the company reportedly declared an internal “code red” over intensifying competition from Google and Anthropic.

Positioned as OpenAI’s most advanced enterprise-focused model yet, GPT-5.2 is targeted at users who rely on AI for daily professional workloads rather than just quick queries.

What Is GPT-5.2?

GPT-5.2 marks the next step in OpenAI’s AI model evolution, promising faster, more accurate, and more reliable results across a wide range of workplace tasks. From reading extremely long documents to generating spreadsheets, creating presentations, writing code, debugging and managing multi-step workflows, the new model moves closer to the performance of a human expert.

OpenAI says enterprise customers using ChatGPT have already reported saving 40–60 minutes per day—and GPT-5.2 is expected to push productivity gains even further.

Three Versions for ChatGPT Users

GPT-5.2 is available in ChatGPT across three modes:

• Instant: Quick responses for everyday tasks
• Thinking: Improved structured reasoning for complex problems
• Pro: Highest-quality outputs for technical and advanced tasks

The model is rolling out first to paid ChatGPT plans. On the API, it is live as gpt-5.2, gpt-5.2-chat-latest, and gpt-5.2-pro. Despite higher token rates than GPT-5.1, OpenAI says improved efficiency results in lower overall costs for many tasks.

How GPT-5.2 Performs

Excels in Professional Benchmarks

The model underwent GDPval—one of the largest tests evaluating performance across 44 real-world professions. GPT-5.2’s “Thinking” variant matched or outperformed industry professionals in 70.9% of tasks, nearly doubling GPT-5’s performance.

Record-Setting Coding Performance

On SWE-Bench Pro, a benchmark simulating real engineering workflows, GPT-5.2 achieved the best results to date. It performed strongly across:

• Debugging
• Reviewing code
• Implementing new features
• Handling full end-to-end engineering tasks
• Developers also noted major improvements in generating 3D UIs, complex visuals and front-end applications.

Massive Long-Context Ability

One of GPT-5.2’s standout upgrades is its ability to process hundreds of thousands of tokens with high accuracy. It maintained near-perfect recall even when critical details were buried deep inside enormous documents—ideal for legal contracts, research papers, transcripts and multi-file projects.

Better Multi-Step Tool Use

The model demonstrated 98.7% accuracy on Tau2 benchmarks for telecom customer-support workflows. It can execute multi-tool, multi-step tasks—such as rebooking travel, locating lost luggage, arranging hotels, or handling medical-seating requests—without losing context.

Sharper Vision and Scientific Reasoning

GPT-5.2 also shows major improvements in visual understanding, including:

• Charts and dashboards
• Software interface screenshots
• Technical diagrams
• Low-resolution images

On graduate-level science problems, it crossed 92% accuracy, and it set new records on expert-level mathematical reasoning tests.

Competitive Pressure Rising

The release of GPT-5.2 follows Google’s strong performance with Gemini 3 and Anthropic’s launch of Claude Opus 4. OpenAI CEO Sam Altman reportedly urged teams to focus on improving model quality while delaying other initiatives, including ad integrations.

With GPT-5.2, OpenAI aims to deliver more economic value across workplaces by enhancing productivity, reducing time spent on repetitive tasks, and offering a more reliable AI partner for professional workloads.