A 77.1% score on the ARC-AGI-2 benchmark marks Google's latest AI model as more than twice as capable at logical reasoning than its predecessor. Gemini 3.1 Pro launched this week with what the company calls "a step forward in core reasoning," positioning it for complex problem-solving tasks where simple answers fall short.
The model achieves its dramatic performance leap by scoring 77.1% on the ARC-AGI-2 test, which evaluates how well AI systems solve entirely new logic patterns they've never encountered before. That figure represents more than double the reasoning capability of Gemini 3 Pro, which Google released just three months earlier in November.
Google is making the upgraded intelligence available immediately through multiple channels. Users can access Gemini 3.1 Pro via the Gemini app and NotebookLM, while developers get preview access through Vertex AI and the Gemini API in Google AI Studio.
For consumers, the model arrives in preview with full general availability expected soon. Subscribers to Google AI Pro and Ultra plans receive higher usage limits within the Gemini app and NotebookLM environment.
The timing matches Apple's recent announcement of a multi-year partnership to use Gemini as the backbone for its next-generation Siri assistant. According to reports, Apple plans to debut Gemini-powered Siri features via iOS 26.4, though recent reports suggest some features may be delayed while Apple confirms the updated Siri is "still on track to launch in 2026."
Google frames Gemini 3.1 Pro as built specifically for scenarios "where a simple answer isn't enough." The company says it handles demanding queries and multi-layered tasks, generating visual explanations of complex topics and synthesizing large data volumes into unified perspectives.
This release follows last week's major update to Gemini Deep Think, which targeted scientific and research challenges across multiple disciplines.
