Google launches Gemini 3.1 Pro for complex tasks

Google has launched Gemini 3.1 Pro, an upgraded version of its artificial intelligence model for solving tasks that require more reasoning than a quick answer. The model is arriving in preview for developers and businesses, and is beginning to roll out to users of Google's consumer products.

The update focuses on solving new and complex problems. In the ARC-AGI-2 test, which measures whether a model can independently discover logical patterns it has not seen before, Gemini 3.1 Pro achieved a verified score of 77.1%. According to Google, that is more than double the reasoning performance of Gemini 3 Pro.

What you can do with Gemini 3.1 Pro

Google is positioning the model for situations where a simple answer is not enough. For example, it can help you:

• Explain a complicated topic visually.
• Bring scattered data together in a single view.
• Work on creative projects with more detailed instructions.
• Solve technical problems that require multiple steps.

The difference is not just that it can provide more information, but that it can better analyze a request before reaching a solution. In practice, that could help you turn extensive information into a useful summary, design a strategy based on several documents, or build software with less manual intervention.

Where it is available

The rollout covers Google's products for different types of users:

• Developers: Gemini API in Google AI Studio, Gemini CLI, Google Antigravity, and Android Studio.
• Businesses: Vertex AI and Gemini Enterprise.
• Users: the Gemini app and NotebookLM.

In the Gemini app, the model is being enabled with higher usage limits for people on the Google AI Pro and Ultra plans. In NotebookLM, it is available exclusively to subscribers of those two plans.

For developers and businesses, Gemini 3.1 Pro remains in preview. That means Google is testing it with real users before offering a general release, while continuing to work on areas such as agentic workflows, which can execute several actions in sequence.

The improvement comes just a few months after Gemini 3 Pro, launched in November. For you, the change will depend on which tool you use and your plan, but the direction is clear: Google's models are moving from answering questions to collaborating on long, technical, multi-step tasks. The important thing now is to see whether that benchmark advantage also holds up in everyday work.