Google launches Gemini 3.1 Pro with stronger reasoning and complex task handling

Smarter AI for complex workflows

Google launches Gemini 3.1 Pro with stronger reasoning and complex task handling

Google has introduced the Gemini 3.1 Pro AI model, marking a significant update to its Gemini ecosystem. The new release focuses on improved reasoning, deeper problem-solving, and enhanced intelligence designed to power more demanding workflows.

With the rollout now underway across consumer and enterprise platforms, the company is positioning this upgrade as a foundation for future AI experiences.

Gemini 3.1 Pro brings stronger reasoning

According to Google, Gemini 3.1 Pro represents a meaningful improvement over earlier Gemini 3 models by refining core intelligence rather than simply adding new features. The model is intended for situations where straightforward responses are insufficient, such as explaining complicated concepts visually, combining large datasets into cohesive outputs, or supporting creative and technical projects.

Google Pixel 10a 5G (8GB RAM, 256GB, Lavender)

Buy now

Google Pixel 10 5G (12GB RAM, 256GB, Obsidian)

Buy now

It is now available through the Gemini app, NotebookLM for select subscribers, developer tools such as Google AI Studio, and enterprise platforms including Vertex AI. Sundar Pichai highlighted that the upgraded baseline allows users to tackle highly complex tasks more efficiently, while Google Labs executive Josh Woodward pointed to improvements in agentic workflows, coding, and structured data synthesis.

Benchmark scores and real-world positioning

Google also emphasised performance gains across several industry benchmarks. Gemini 3.1 Pro reportedly achieved a verified score of 77.1 percent on the ARC-AGI-2 test, which measures a model’s ability to solve unfamiliar logic problems, marking a substantial leap compared to the previous version.

On Humanity’s Last Exam, it secured 44.4 percent, ahead of competing models in standard evaluation scenarios, though some rivals still maintain a narrow edge when additional tools are enabled. In software engineering benchmarks such as SWE-Bench Verified, results remain closely contested among leading AI systems.

The company says this preview phase will help refine agent-based workflows further before broader availability, signalling continued rapid iteration in the evolving AI model landscape.

ALSO READ: Google adds Nano Banana to Chrome with deeper Gemini integration

Stay tuned to Unboxed by Croma for all the updates about the AI world.

Unleash your inner geek with Croma Unboxed

Subscribe now to stay ahead with the latest articles and updates

You are almost there

Enter your details to subscribe

0

Disclaimer: This post as well as the layout and design on this website are protected under Indian intellectual property laws, including the Copyright Act, 1957 and the Trade Marks Act, 1999 and is the property of Infiniti Retail Limited (Croma). Using, copying (in full or in part), adapting or altering this post or any other material from Croma’s website is expressly prohibited without prior written permission from Croma. For permission to use the content on the Croma’s website, please connect on contactunboxed@croma.com

Comments

Leave a Reply
  • Related articles
  • Popular articles
  • Gaming

    GTA V cheat codes: A complete list

    Karthekayan Iyer

  • Gaming

    GTA San Andreas cheats and codes

    Shubhendu Vatsa

  • Smartphones

    All Apple iPhones launched since 2007

    Chetan Nayak