Google DeepMind has unveiled Gemini 3.1 Pro, a significant upgrade designed to handle tasks where simple answers fall short. While last week's update focused on specialized scientific breakthroughs, 3.1 Pro brings that advanced logic to everyday applications. It is currently rolling out across the Gemini API, Google AI Studio, and consumer apps like NotebookLM, providing a smarter baseline for complex challenges.

The model's standout feature is its performance on the ARC-AGI-2 benchmark. This test measures a model's ability to solve entirely new logic patterns it hasn't seen during training, rather than just memorizing data. With a verified score of 77.1%, Gemini 3.1 Pro more than doubles the performance of the previous 3 Pro version, marking a massive leap in abstract problem-solving.

Practical applications of this intelligence are diverse. The model can generate website-ready animated SVGs—scalable graphics built with code—and synthesize complex data into live dashboards, such as visualizing the International Space Station’s orbit. It also demonstrates creative coding by translating literary themes from novels like Wuthering Heights into functional, aesthetically matched user interfaces.

For developers, the release includes higher usage limits for Google AI Pro and Ultra subscribers. This preview phase aims to refine performance in multi-step agentic workflows, where the AI manages complex processes autonomously, before a broader general release later this year.

Google DeepMind has unveiled Gemini 3.1 Pro, a significant upgrade designed to handle tasks where simple answers fall short. While last week's update focused on specialized scientific breakthroughs, 3.1 Pro brings that advanced logic to everyday applications. It is currently rolling out across the Gemini API, Google AI Studio, and consumer apps like NotebookLM, providing a smarter baseline for complex challenges.

The model's standout feature is its performance on the ARC-AGI-2 benchmark. This test measures a model's ability to solve entirely new logic patterns it hasn't seen during training, rather than just memorizing data. With a verified score of 77.1%, Gemini 3.1 Pro more than doubles the performance of the previous 3 Pro version, marking a massive leap in abstract problem-solving.

Practical applications of this intelligence are diverse. The model can generate website-ready animated SVGs—scalable graphics built with code—and synthesize complex data into live dashboards, such as visualizing the International Space Station’s orbit. It also demonstrates creative coding by translating literary themes from novels like Wuthering Heights into functional, aesthetically matched user interfaces.

For developers, the release includes higher usage limits for Google AI Pro and Ultra subscribers. This preview phase aims to refine performance in multi-step agentic workflows, where the AI manages complex processes autonomously, before a broader general release later this year.