Google DeepMind released Gemini 3.1 Pro on February 19, marking a major leap in AI reasoning. The model scored 77.1% on the ARC-AGI-2 benchmark — more than double the score of Gemini 3 Pro released just three months earlier. It features a 1-million-token context window and can output up to 64,000 tokens, processing text, images, audio, video, and code simultaneously.
On practical engineering tasks, Gemini 3.1 Pro cut an optimization challenge runtime from 300 seconds to just 47 seconds, beating the human reference solution of 94 seconds. Google priced it at $2 per million input tokens and $12 per million output tokens — less than half the cost of Anthropic's Claude Opus 4.6 at $5/$25.
The model is rolling out across the Gemini app, NotebookLM, Google AI Studio, Vertex AI, and Android Studio. Google described it as designed for "tasks where a simple answer isn't enough," with upgraded reasoning inherited from Gemini 3 Deep Think.
