The release comes as a result of the “Code Red” mode declared by CEO Sam Altman in early December.
According to the company, GPT-5.2 significantly outperforms previous versions in tasks such as information retrieval, text translation, coding, and image processing. In addition, the new model is capable of generating spreadsheets, preparing presentations, retaining long texts, and solving complex, multi-step tasks.
The model’s performance is also backed by numbers. In “GDPval” tests covering 44 different fields of intellectual work, GPT-5.2 matched or surpassed human experts in 70.9% of cases — a substantial leap from the 38.8% achieved by GPT-5.1. Programming test results have also improved by roughly 5 percentage points.
Currently, the “Instant,” “Thinking,” and “Pro” versions of the model are available to ChatGPT paid subscribers and API users.
This rapid update was driven by Google’s recent breakthrough. Earlier in December, The Information reported that Google’s new Gemini 3 Pro model had outperformed GPT-5.1 in OpenAI’s internal tests, prompting Sam Altman to declare an emergency within the company. Several other projects were put on hold, and the entire team was mobilized to enhance ChatGPT.
