Gemini 3.1 Pro: Google's Upgraded AI Model Doubles Down on Reasoning

• 2Fifteen Tech
Google Workspace AI Productivity
Google Gemini

Three months after releasing Gemini 3 Pro, Google is already shipping a significant upgrade. Gemini 3.1 Pro is rolling out across the Gemini app, NotebookLM, and developer platforms with one headline improvement: substantially better reasoning.

What Changed

Gemini 3.1 Pro takes the intelligence behind last week’s Gemini 3 Deep Think model and brings it into the standard Gemini experience. Better reasoning without needing to toggle into a separate mode.

The benchmarks reflect this. On ARC-AGI-2, which tests a model’s ability to solve novel logic patterns, 3.1 Pro scored 77.1%, more than double the 31.1% from Gemini 3 Pro. On Humanity’s Last Exam, it set a new record at 44.4%, up from 37.5%.

In practice, this means stronger performance on tasks where a simple lookup won’t cut it: synthesizing information from multiple sources, multi-step problem solving, nuanced data analysis, and more sophisticated code and creative output.

Where to Access It

3.1 Pro is rolling out today in preview:

  • Gemini app - Higher usage limits for Google AI Pro and Ultra subscribers
  • NotebookLM - Exclusively for Pro and Ultra users
  • Google AI Studio, Gemini CLI, Vertex AI, Gemini Enterprise - For developers and enterprises
  • Android Studio and Google Antigravity - For mobile and agentic development

API pricing is unchanged ($2/$12 per million input/output tokens) and the context window stays at 1 million input tokens.

Where You’ll Notice the Difference

Complex analysis. Pulling together information from multiple documents, comparing options, or working through multi-variable problems should produce more thorough results.

Agentic workflows. The model nearly doubled its score on agentic task benchmarks. Multi-step automations where the AI makes decisions and takes actions should be more reliable.

Creative and technical output. Google highlighted improved code-based visualizations, interactive designs, and complex simulations. First-pass quality for coding and prototyping should be noticeably better.

For everyday Gemini app conversations, the improvements will be subtler. You’ll notice the difference most when asking harder questions.

How It Compares

Gemini 3.1 Pro leads on several benchmarks, particularly ARC-AGI-2 and Humanity’s Last Exam. On community-driven evaluations like the Arena leaderboard, it’s competitive with but doesn’t clearly lead other frontier models.

If you’re already in the Google ecosystem, 3.1 Pro is a meaningful upgrade at no additional cost to your existing subscription. That alone makes it worth trying.

What This Means for Google Workspace Users

Each Gemini model upgrade makes the AI features across Gmail, Docs, Sheets, Meet, and the Gemini app more capable without requiring any changes on your end. Google is shipping these improvements more frequently now, which means the tools you’re already paying for keep getting better.

Gemini 3.1 Pro isn’t a revolutionary new product. It’s a meaningful step forward in intelligence for a model millions of people use daily, available at the same price point as its predecessor. If you use the Gemini app or NotebookLM, you’ll start seeing it roll out over the coming days.

Need Help With Google Workspace?

At 2Fifteen Tech, we help businesses get the most out of Google Workspace, including staying current with AI capabilities like Gemini. If you want help configuring your environment or understanding which features make sense for your team, we can help.

Back to Blog