OpenAI releases GPT-5.2 to take on Google and Anthropic (www.engadget.com)

OpenAI has launched GPT-5.2 as a direct response to the competitive pressures from Google's Gemini 3 Pro, emphasizing it as their most advanced model yet for practical professional applications. The new version boasts significant improvements in creating spreadsheets, presentations, code writing, and complex project management. In benchmark tests such as AIME 2025 and ARC-AGI-1, GPT-5.2 demonstrated superior performance over its predecessor, achieving a perfect score in mathematics problems and showing enhanced abstract reasoning capabilities. Notably, OpenAI claims GPT-5.2 reduces factual error rates by 30%, increasing its reliability for tasks like research and decision support. This release is pivotal for OpenAI as the company faced criticism for the previous GPT-5 model, which failed to meet user expectations and was perceived as underwhelming. The introduction of three variants—Instant, Thinking, and Pro—aims to cater to diverse user needs, with the Instant version designed for everyday tasks and improved conversational interactions. The pressure is on OpenAI to ensure GPT-5.2 can compete effectively against rival models, especially as it rolls out amidst high stakes with substantial business investments in infrastructure.