Google Ships Gemini 3.1 Pro, More Than Doubling Reasoning Score

Google has released Gemini 3.1 Pro in preview, and the model more than doubles its predecessor's score on ARC-AGI-2, a benchmark that tests a system's ability to solve entirely new logic patterns. The verified score of 77.1% is, according to Google, more than double the reasoning performance of Gemini 3 Pro, released in November.

The release lands one week after Google shipped a major update to Gemini 3 Deep Think, its model aimed at science, research and engineering workloads. In its announcement, Google framed 3.1 Pro as the upgraded core intelligence behind those capabilities. "Today, we're releasing the upgraded core intelligence that makes those breakthroughs possible: Gemini 3.1 Pro," the company said.

The rollout is broad. Starting today, developers can access 3.1 Pro in preview through the Gemini API in Google AI Studio, Gemini CLI, the agentic development platform Google Antigravity, and Android Studio. Enterprise customers get the model through Vertex AI and Gemini Enterprise. Consumers see it in the Gemini app, where it ships with higher usage limits for subscribers on the Google AI Pro and Ultra plans, and in NotebookLM, where it is available exclusively to Pro and Ultra users.

What the benchmark means

ARC-AGI-2 matters because it measures something distinct from recall or pattern-matching on known data: the ability to reason through novel logic problems a model has not encountered before. A jump from 3 Pro's level to 77.1% suggests a genuine step in generalization rather than incremental tuning. Google calls 3.1 Pro "a smarter, more capable baseline for complex problem-solving."

The company positions the model for tasks where a single correct answer is insufficient — synthesizing data into a unified view, generating clear visual explanations of complex topics, or executing multi-stage creative projects. Those use cases are precisely where AI vendors are competing for enterprise spending, and where agentic workflows have become the industry's stated next frontier.

Preview first, then general availability

Google is explicit about why the model ships in preview rather than as a generally available release. "We are releasing 3.1 Pro in preview today to validate these updates and continue to make further advancements in areas such as ambitious agentic workflows before we make it make it generally available soon," the company said. The phrasing signals that Google wants real-world feedback before committing to a stable release — the same dynamic the company credits for the fast iteration since Gemini 3 Pro launched in November.

"Since releasing Gemini 3 Pro in November, your feedback and the pace of progress have driven these rapid improvements," Google said.

The cadence itself is notable. Two major model releases in roughly a month — the Deep Think update last week and 3.1 Pro today — indicate Google is compressing its release cycle in competition with OpenAI and Anthropic, both of which have accelerated their own shipping schedules this year.

For developers and enterprises evaluating foundation models, the practical takeaway is immediate: 3.1 Pro is available now across Google's API, IDE tooling and cloud platforms at preview status, with general availability promised soon. For consumers, access depends on subscription tier — the highest-capability version of the model sits behind the Google AI Pro and Ultra paywalls in both the Gemini app and NotebookLM.