Google unveils Gemini 3.1 Pro with a focus on advanced logical reasoning

In November, Google rolled out an early version of Gemini 3 Pro, followed a month later by Gemini 3 Flash. Now, the company has officially announced Gemini 3.1 Pro, a model designed for challenges where a simple answer just won’t cut it.

This .1 update marks a first for Google. Previous Gemini releases debuted intermediate versions labeled .5-like Gemini 2.5 Pro, which was first previewed in March and refined ahead of the I/O conference in May.

According to Google, Gemini 3.1 Pro takes a significant step forward in foundational reasoning capabilities. The upgraded “intelligence core,” which debuted last week in Gemini 3 Deep Think, is now widely available as part of the 3.1 Pro release.

The new model clocks a 77.1% score on the ARC-AGI-2 benchmark-more than double the logical reasoning performance of Gemini 3 Pro. This advancement in “advanced reasoning” translates into the ability to clearly explain complex topics, synthesize data into a coherent whole, and assist with creative projects.

Gemini 3.1 Pro is tailored for complex problems that require more than just answers-it’s built for deep analysis and generating insightful conclusions to tackle truly difficult tasks.

The model is already being distributed through the Gemini app and is available within NotebookLM for Google AI Pro and Ultra subscribers. Additionally, Gemini 3.1 Pro is accessible via the Gemini API through Google AI Studio, Antigravity, Vertex AI, Gemini Enterprise, Gemini CLI, and Android Studio for developers.

Currently, version 3.1 Pro is launched as a preview release. This phased rollout allows Google to test real-world applications and continue expanding capabilities-like complex agent scenarios-before the model becomes publicly available soon.