OpenAI Officially Introduces its GPT-5.2 Flagship Model: Instant, Thinking, and Pro Versions Rolling Out
The American AI tech brand – OpenAI has introduced its GPT-5.2 flagship model following the previous GPT-5.1 model. It has already been made available to a few of the paid users, and later will reach other users as well.
Check out more about the latest GPT-5.2 model and its features below.
GPT-5.2 Flagship Model – Officially Introduced
On the 11th of December 2025 (Thursday), OpenAI officially introduced its next GPT-5.2 flagship model, and the Instant, Thinking, and Pro versions are currently rolling out. Initially, these GPT-5.2 versions will be made accessible for users of the Plus, Pro, Business, and Enterprise plans, and later will be made available to the Free and Go users as well. Notably, the brand has also made it available on the API and Codex too.
In terms of performance, when compared to the previous GPT-5.1 (Thinking version), the GPT-5.2 (Thinking version) has attained higher performance scores on the SWE-Bench Pro (55.6%), GPQA Diamond (92.4%), CharXiv Reasoning (82.1%), FrontierMath (40.3%; 14.6%), AIME 2025 (100.0%), ARC-AGI-1 (86.2%), ARC-AGI-2 (52.9%), and GDPval (70.9%) benchmarks. Speaking more, on the GDPval benchmark, the GPT-5.2 (Thinking version) has become the very first model to perform at human expert level in tasks like making presentations, spreadsheets, and more.
Key features which include the state-of-the-art long-context reasoning, improvements in productivity tasks (spreadsheet creation, analysis, and formatting), and early gains in slideshow creation are a few of its notable aspects. Also, the Instant version of the newly-unveiled GPT-5.2 model has been built for everyday work and learning, and it assures a warm and conversational experience just like the GPT-5.1 model. Clearer explanations, improved how-tos and walk-throughs, stronger technical writing and translation, and better studying and career guidance support are its further aspects. About the GPT-5.2 Pro version, it is claimed to be the smartest and most trustworthy model for difficult questions, with stronger performance in domains like programming and best assistance for scientists.
OpenAI has also added that the latest GPT-5.2 flagship model is a part of an ongoing set of model improvements, and the brand aims in continuing its work to address known issues such as over-refusals and latency. To note, while the GPT-5.2 is now out, ChatGPT will still be offering the GPT-5.1 model to its paid users as its legacy model for a duration of three months.
