Google has launched Gemini 3 Deep Think, its most advanced reasoning mode designed for AI Ultra subscribers within the Gemini app. This upgrade enhances AI’s ability to tackle complex, multi-step problems in diverse fields like science and mathematics. Deep Think allows parallel reasoning, evaluating multiple hypotheses simultaneously, resulting in more precise and structured outcomes. In tests, Deep Think achieved a remarkable 41.0% on “Humanity’s Last Exam,” outpacing previous models and marked a milestone by reaching 45.1% on ARC-AGI-2 when combined with code execution. Google emphasizes that Deep Think goes beyond basic conversational AI, offering a robust analytical tool for researchers and professionals. Access to this new mode is immediate for Ultra subscribers via the Gemini app. This launch positions Google strategically in the competitive AI landscape, as it aims to redefine machine reasoning and address real-world challenges, setting itself apart from competitors like OpenAI and Anthropic.
Source link
Share
Read more