Google releases Gemini 3.1 Pro Thinking. Reasoning capabilities take a quantum leap: The ARC-AGI-2 benchmark score is more than double that of Gemini 3.0 Pro – meaning it’s no longer just mimicking when handling unseen complex logic and brand-new problems, but now possesses deeper thinking abilities. Transforming AI from a text generator to a system builder. Super excited – can it outperform Claude Opus 4.6 on OpenClaw? https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/