In December 2025, OpenAI unveiled the GPT - 5.2 series models. This series comprises three versions: Instant, Thinking, and Pro. The launch was aimed at taking on Google's Gemini 3 Pro in the competitive AI landscape. GPT - 5.2 showcased remarkable performance in a number of benchmark tests, especially excelling in professional knowledge - based work tasks. The GPT - 5.2 Thinking version outperformed or was on par with industry experts in 70.9% of the tasks. Moreover, it generated results 11 times faster than human experts, all while incurring a cost of less than 1% of what human experts would require. Substantial enhancements were also evident in areas such as mathematics, coding, long - document processing, and visual capabilities. Notably, it achieved a 38% reduction in hallucination rates compared to its predecessor model. However, despite setting records in multiple tests, Google's Gemini 3 Pro still held the upper hand in some benchmark tests. For instance, it scored 46.4% in the Humanity’s Last Exam test. OpenAI asserts that GPT - 5.2 is its most powerful model to date, specifically engineered to generate economic value. Nevertheless, whether it can reclaim market leadership is a question that only time will answer.