Introducing Google’s Gemini 3.1 Pro: A Leap Forward in AI Reasoning Performance
On February 19, 2026, Google unveiled the Gemini 3.1 Pro, marking the company’s inaugural .1 increment release. This launch signifies what Google terms as a “step forward in core reasoning” for its AI models, particularly designed for intricate problem-solving tasks where simplistic answers fall short.
The Gemini 3.1 Pro has achieved a remarkable 77.1% score on the ARC-AGI-2 benchmark. This score is more than double the reasoning performance of its predecessor, the Gemini 3 Pro, which scored 31.1%. This significant enhancement in advanced reasoning has practical implications for tasks that require data synthesis, visual explanations of complex subjects, and the realization of creative projects.
Starting today, the model is rolling out across Google’s ecosystem. For consumers, it’s accessible in the Gemini app for all users, with higher limits set for Google AI Pro and Ultra subscribers. It’s also exclusively available in NotebookLM for Pro and Ultra users. Developers and enterprises can avail Gemini 3.1 Pro through the Gemini API via Google AI Studio, Vertex AI, Android Studio, and other platforms.
Google is releasing this version in preview to validate updates and advance agentic workflows – systems where AI can take on multi-step tasks with more independence. The company’s release represents its commitment to bringing breakthrough intelligence to everyday applications. CEO Sundar Pichai noted that Gemini now processes over 10 billion tokens per minute via direct API use.
Source: Google Official Blog