This month, Anthropic launched versions Opus and Sonnet of Claude 4.6, outperforming Google’s Gemini 3 Pro in various aspects. In response, Google introduced the advanced Gemini 3.1 Pro, which boasts enhanced reasoning capabilities, achieving a benchmark score of 77.1% on the ARC-AGI-2, significantly surpassing its predecessor. Though 3.1 Pro’s multimodal understanding dropped slightly to 80.5%, it still displays notable upgrades in agentic coding, nearly matching Claude’s performance. Google emphasizes practical applications for 3.1 Pro, including generating animated SVGs and visual explanations, indicating a focus on user experience. The rollout commenced today across multiple platforms, including the Gemini app and NotebookLM, with exclusive access for Pro and Ultra users. Google’s commitment to continuous improvement is evident through user feedback, aiming for rapid iterations in AI models before wider availability. Overall, Gemini 3.1 Pro enhances AI capabilities while maintaining competitive edge against Anthropic’s Claude models.
Source link
