Google's Gemini 3.1 Pro launches with record benchmark scores, showcasing advanced reasoning capabilities for complex tasks.
Google has introduced Gemini 3.1 Pro, a powerful new iteration of its AI model that has achieved record-breaking results in various benchmarks.
This model stands out in the competitive AI landscape.
According to the announcement, the upgrade features improved core reasoning capabilities, transforming it into a more intelligent neural network adept at handling intricate tasks.
In the ARC-AGI-2 test, which assesses the model's logical reasoning, Gemini 3.1 Pro recorded a score of 77.1%, significantly outperforming its predecessor, Gemini 3 Pro, by more than twofold.
Gemini 3.1 Pro is currently accessible to:
- developers in a preview phase via the Gemini API, Google AI Studio, and Android Studio;
- enterprise users through the Vertex AI and Gemini Enterprise platforms;
- the public via the Gemini app and NotebookLM.
The developers emphasized that Gemini 3.1 Pro excels in situations where merely providing a straightforward answer isn't sufficient, utilizing advanced reasoning to tackle the most challenging problems.
They also highlighted that the enhanced capabilities are beneficial in various practical contexts, including simplifying complex topics, synthesizing information, and developing creative projects.
Earlier this year, Google enhanced the reasoning functionality in Gemini 3 Deep Think, aimed at addressing complex tasks in scientific and engineering domains.
