Google Releases Gemini 3.1 Pro With Leading Reasoning Benchmark Scores

Summary

Google DeepMind released Gemini 3.1 Pro on February 19, posting leading scores on 13 of 16 benchmarks. The headline result is 77.1% on ARC-AGI-2 — a test of pure logic and novel problem-solving — more than double the reasoning performance of the standard Gemini 3 Pro. The model features a 1M token context window and enhanced multimodal capabilities spanning text, audio, images, video, and entire code repositories. Gemini 3.1 Pro is available on Vertex AI, Gemini Enterprise, Gemini CLI, Android