Sou Hu Cai Jing·2026-02-20 12:39Core Insights - Google has officially launched its latest AI model, Gemini 3.1 Pro, which has doubled its reasoning capabilities compared to its predecessor, achieving a score of 77.1% on the ARC-AGI-2 benchmark, marking a significant breakthrough in AI technology [3][4]. Group 1: Model Performance - Gemini 3.1 Pro shows a notable improvement in reasoning ability, particularly excelling in handling new logical patterns, indicating major advancements in Google's reasoning capabilities [3]. - Since its release in November last year, Gemini 3 has performed exceptionally well in various internal task tests, surpassing several competitors, including Microsoft's Copilot, and has received high user ratings, laying a solid foundation for Gemini 3.1 Pro [3]. Group 2: Recent Upgrades - The release of Gemini 3.1 Pro follows a significant upgrade to Gemini 3 Deep Think, which introduced new capabilities in chemistry and physics, as well as breakthroughs in mathematics and coding [4]. - Google aims to address complex research challenges with the upgrades, positioning Gemini 3.1 Pro as a core intelligence that facilitates new breakthroughs [4]. Group 3: Competitive Landscape - While Gemini 3.1 Pro has made progress in reasoning capabilities, Anthropic's Claude Opus 4.6 remains at the top of the text capability rankings, showcasing its advantages in reasoning and safety [5]. - The introduction of Gemini 3.1 Pro adds new competitive dynamics to the market, with the AI model competition expected to intensify as new models like GPT-5.3 emerge [5]. - The lifecycle of AI models extends beyond a single release, involving continuous testing and iteration for optimization, posing a challenge for Google to maintain its competitive edge [5].