OpenAI has announced the release of GPT-5.2, the newest upgrade in its GPT-5 series, designed to handle professional knowledge work, advanced reasoning, and complex agent-based tasks. The rollout has begun immediately, with early access granted to paid users, marking a major advancement in both performance and safety.
GPT-5.2 is not a single model but a family of models, including GPT-5.2 Instant, GPT-5.2 Thinking, and GPT-5.2 Pro. OpenAI says the new series delivers significant improvements in mathematics, science, coding, imaging, agentic workflows, and overall accuracy, making it the most capable GPT release so far for professional use.
One of the key upgrades is a reduction in hallucinations. According to OpenAI, GPT-5.2 Thinking shows an average hallucination rate of 10.9%, down from 12.7% in GPT-5.1 Thinking. When paired with web access, the rate drops further to 5.8%, increasing reliability for real-world applications.
The model shows strong gains across professional workflows, including spreadsheets, presentations, and autonomous task execution. OpenAI says GPT-5.2 performs at or above human expert levels on GDPval, a benchmark covering 44 occupations. Output generation is reportedly up to 11 times faster than before and costs less than 1% of what comparable human expertise would require.
In software development, GPT-5.2 sets a new benchmark by scoring 55.6% on SWE-Bench Pro, which evaluates real-world software engineering tasks across multiple programming languages. The model also demonstrates improved performance in interactive coding, code reviews, and bug detection, based on internal and third-party testing.
GPT-5.2 also delivers stronger results in scientific reasoning, advanced mathematics, abstract problem-solving, and computer vision. Enhanced tools allow the model to handle complex, multi-step tasks more efficiently than earlier versions.
OpenAI has placed a stronger emphasis on safety in this release, particularly in conversations involving mental health, suicide risk, and emotional dependency. Internal evaluations suggest GPT-5.2 produces fewer harmful or inappropriate responses than GPT-5.1. The update follows earlier criticism and legal challenges related to AI behavior in sensitive situations, with OpenAI stating that targeted safeguards have been added to reduce risk.
GPT-5.2 is now available to Plus, Pro, Go, Business, and Enterprise subscribers, with a phased rollout to maintain stability. Older GPT-5.1 models will remain accessible for the next three months. Developers can also access GPT-5.2 through the API, enabling integration into professional, research, and agent-driven applications.
