Deep thinking AI has seen groundbreaking advancements with Google Gemini 3 and DeepSeek models, revolutionizing artificial intelligence’s problem-solving capabilities. Google Gemini 3 Pro, launched on November 18, 2025, achieved an impressive score of 1501 Elo on the LMArena Leaderboard, surpassing prior models. The model excels in reasoning, scoring 41.0% on Humanity’s Last Exam without tools and 45.1% on ARC-AGI-2 with code execution. Meanwhile, DeepSeek’s V3.2 models launched on December 1, 2025, with the Speciale variant achieving a 96.0% score in mathematics competitions. Gemini 3’s Deep Think mode harnesses advanced parallel reasoning and is exclusive to Google AI Ultra subscribers, pushing reasoning to unprecedented levels. This dual breakthrough from Google and DeepSeek emphasizes a shift towards authentic reasoning over speed, enabling complex problem-solving across industries like scientific research and software development. With this evolution, enterprises can now access AI that thinks critically, marking a pivotal moment in AI’s future.
Source link
