The frontier of professional AI has just taken a significant leap forward. OpenAI has introduced GPT-5.2, a model series designed to unlock unprecedented economic value in knowledge work. Already, the average ChatGPT Enterprise user reports saving 40–60 minutes daily, with heavy users reclaiming over 10 hours a week. GPT-5.2 is engineered to build on this, offering superior capabilities in creating spreadsheets, building presentations, writing code, perceiving images, understanding long contexts, using tools, and handling complex, multi-step projects.
A New Benchmark in Professional Performance
GPT-5.2 sets a new state of the art across numerous benchmarks. Most notably, on GDPval—an evaluation measuring well-specified knowledge work tasks across 44 occupations—GPT-5.2 Thinking beats or ties top industry professionals on 70.9% of comparisons. This is a substantial jump from the previous generation. The model produced outputs for these tasks at over 11x the speed and less than 1% of the cost of expert professionals, suggesting a powerful new tool for professional workflows under human oversight.
Example of a sophisticated workforce planning model generated by GPT-5.2 Thinking.
Supercharged Coding and Development
For software engineers, GPT-5.2 Thinking is a game-changer. It achieves a new state-of-the-art score of 55.6% on SWE-Bench Pro, a rigorous, multi-language evaluation of real-world software engineering. This translates to a model that can more reliably debug production code, implement feature requests, refactor large codebases, and ship fixes end-to-end with less manual intervention. Early testers found it significantly stronger at front-end development and complex UI work, especially involving 3D elements.
Enhanced Reliability and Long-Context Mastery
Factuality sees a meaningful improvement, with responses containing errors becoming 30% less common compared to GPT-5.1 Thinking. For long-context reasoning, GPT-5.2 Thinking achieves leading performance, enabling professionals to work with reports, contracts, and multi-file projects across hundreds of thousands of tokens while maintaining coherence and accuracy. It’s the first model to achieve near 100% accuracy on challenging 4-needle retrieval tasks out to 256k tokens.
Sharper Vision and Robust Tool Calling
GPT-5.2 is the strongest vision model yet from OpenAI, cutting error rates roughly in half on chart reasoning and software interface understanding. This allows for more accurate interpretation of dashboards, technical diagrams, and visual reports. Its spatial understanding is markedly improved, as shown in its ability to identify and locate components within a complex image like a motherboard.
GPT-5.1’s attempt at labeling motherboard components.
GPT-5.2 shows a significantly stronger grasp of spatial arrangement and component identification.
In tool calling, it achieves a new state of the art of 98.7% on Tau2-bench Telecom, demonstrating exceptional reliability in orchestrating long, multi-turn workflows. This is crucial for automating complex customer support cases or data analysis pipelines that require pulling information from multiple systems.
Accelerating Science and Mathematics
One of the core hopes for AI is to accelerate scientific discovery. GPT-5.2 Pro and Thinking are positioned as the world’s best models for this task. On the graduate-level GPQA Diamond benchmark, they achieve 93.2% and 92.4% respectively. On FrontierMath, an evaluation of expert-level mathematics, GPT-5.2 Thinking solves 40.3% of problems, a new state of the art. Researchers have already used GPT-5.2 Pro to explore open questions in statistical learning theory, with the model proposing a proof that was subsequently verified by human experts.
Availability and Refined User Experience
GPT-5.2 is rolling out now in ChatGPT for paid plans (Plus, Pro, Go, Business, Enterprise) and is immediately available via the API. The release includes three variants tailored for different needs:
* GPT-5.2 Instant: A fast, capable workhorse for everyday tasks with clearer explanations.
* GPT-5.2 Thinking: Designed for deeper, more complex work with greater polish.
* GPT-5.2 Pro: The smartest and most trustworthy option for difficult questions where answer quality is paramount.
In the API, GPT-5.2 is priced at $1.75 per million input tokens and $14 per million output tokens, with a 90% discount on cached inputs. While priced higher per token than GPT-5.1, its greater token efficiency means the cost to achieve a given level of quality is often lower.
Built on a Foundation of Safety
GPT-5.2 builds on the “safe completion” research introduced with GPT-5. With this release, OpenAI has made targeted improvements to how models respond to prompts indicating mental health distress, emotional reliance, or self-harm, resulting in fewer undesirable responses. The company is also in the early stages of rolling out an age prediction model to automatically apply content protections for users under 18.
The response from early enterprise testers has been emphatic. AJ Orbach, CEO of Triple Whale, noted: “GPT-5.2 unlocked a complete architecture shift for us. We collapsed a fragile, multi-agent system into a single mega-agent with 20+ tools. The best part is, it just works… It feels like pure magic.”
This release represents a substantial step in making frontier AI intelligence more practical, reliable, and deeply integrated into the fabric of professional work.