Google Gemini 3.1 Pro – the New Benchmark King

Google Gemini 3.1 Pro – the New Benchmark King

Ondřej Barták
Ondřej Barták
Entrepreneur and Programmer
23. 2. 2026
4 minutes reading · 11 views
Google Gemini 3.1 Pro – the New Benchmark King

    The tech world has been shaken once again. Google has quietly but confidently released Gemini 3.1 Pro – a new version of its flagship language model that immediately reshuffled the deck in the battle for the artificial intelligence throne.

    Gemini 3.1

    Gemini 3.1 Pro builds on the foundations of the Gemini 3 series, which Google launched in November 2025 – and which itself received an enthusiastic response at the time as one of the most capable AI tools on the market. But the pace of progress in AI is now so dizzying that three months is enough for a record-holder to become average.

    Google therefore introduced 3.1 Pro as an enhanced intelligence core – a model designed for situations where a simple answer is not enough. Do you want to synthesize complex scientific data into a clear output? Do you need to explain a complicated technical concept in an understandable way? Or do you simply want to build a functional web application from a single text prompt? That is exactly what 3.1 Pro was built for.

    And the results? They are truly impressive.

    Record-breaking benchmarks

    Here comes the part that caused a stir throughout the tech community. On the ARC-AGI-2 benchmark – one of the most demanding tests of an AI's ability to solve entirely new logical patterns that the model has never encountered before – Gemini 3.1 Pro achieved a verified score of 77.1%. That means more than double the performance of its predecessor, Gemini 3 Pro.

    Gemini 3.1 Pro - benchmark results.
    Gemini 3.1 Pro - benchmark results

    It did not stop at ARC-AGI-2. Google shared results from another independent benchmark with the telling name Humanity's Last Exam – and the model did not lag behind there either. The results showed significantly better performance than the previous generation.

    But what is interesting is that the praise did not come from Google alone. Brendan Foody, CEO of the AI startup Mercor, whose APEX evaluation system measures how well AI models handle real professional tasks, wrote on social media: “Gemini 3.1 Pro is now at the top of the APEX-Agents leaderboard” and added that the results show how quickly agentic AI is improving at real-world professional work.

    What can Gemini 3.1 Pro do?

    Numbers are one thing. But what can the model actually do? Google showed several specific examples that are – quite frankly – astonishing.

    Animated SVGs from a text prompt. The model can generate fully animated vector graphics directly from a natural-language description. The result is clean code, not pixels – so the graphics remain sharp at any resolution, and the file is a fraction of the size of a video.

    Live aerospace dashboard. In one demonstration, the model built a functional dashboard tracking the orbit of the International Space Station – configuring a public telemetry stream on its own and visualizing the data in real time. No manual programming required.

    3D starling murmuration with hand tracking. The model programmed a complex 3D simulation of bird flocks in which the user can influence the flock's behavior by moving their hand – and added a generative musical accompaniment that changes according to the birds' movement.

    Wuthering Heights as a modern portfolio. When tasked with creating a modern personal website inspired by Emily Brontë's novel, the model did not limit itself to summarizing the plot. It analyzed the book's atmosphere, narrative tone, and the protagonist's character – and designed an elegant, contemporary website that truly captures that mood.

    Are you wondering whether the average user can do all this, or only an experienced developer? That is precisely where the power of 3.1 Pro lies – it makes advanced reasoning accessible even to those who cannot code.

    How to access Gemini 3.1 Pro

    This time, Google did not focus on just one group of users. The model is being rolled out across the entire ecosystem.

    Developers can try it in preview through the Gemini API in Google AI Studio, as well as through the Gemini CLI, the agentic development platform Google Antigravity, and Android Studio. Businesses and enterprises can find it in Vertex AI and Gemini Enterprise. Regular users can access it through the Gemini app and NotebookLM – although for now, it is available exclusively to subscribers of the Google AI Pro and Ultra plans.

    The battle for the best model continues

    It would be naive to think that Google releases new models every three months just because it enjoys doing so. The AI model race has accelerated dramatically in recent months – OpenAI, Anthropic, and others are constantly releasing new versions of their systems, with each claiming the title of the most capable. And Google cannot afford to fall behind.

    Just a year ago, we were amazed that AI could write a decent email. Today, a model can build a live dashboard tracking the ISS or design a website inspired by a Victorian novel – all from a single sentence.

    Advertisement

    Content created with help from UpTier.

    SEO and GEO on autopilot. UpTier’s multi-agent systems write and optimize content for search engines and AI answers.

    Discover UpTier ↗

    Category:AI
    Did you enjoy this article?
    Discover more interesting posts on our blog
    Back to blog

    Related posts

    OpenAI introduces GPT-6.2 Astra and Sol for professional and everyday workOpenAI introduces GPT-6.2 Astra and Sol for professional and everyday work
    OpenAI has introduced two models: Astra for complex professional tasks and Sol for everyday work. Both are available in ChatGPT Work, Codex and through the API.
    3 min read
    17. 11. 2026
    Tencent releases Hy4 Flash with image support and a context window of up to 1 million tokensTencent releases Hy4 Flash with image support and a context window of up to 1 million tokens
    The lighter Hy4 Flash model targets chat, documents and routine code edits. It accepts text and images, supports tool use and, according to Tencent, matched Claude Haiku 5.5 in tests of everyday tasks.
    1 min read
    16. 11. 2026
    Companies Linked to Chinese Spy Invested in DeepSeek and Promised Banned Nvidia ChipsCompanies Linked to Chinese Spy Invested in DeepSeek and Promised Banned Nvidia Chips
    Companies linked to an alleged Chinese spy joined DeepSeek’s latest funding round. In exchange for a stake, they reportedly promised banned Nvidia Blackwell chips.
    3 min read
    16. 11. 2026
    Přihlaste se k odběru našeho newsletteru
    Zůstaňte informováni o nejnovějších příspěvcích, exkluzivních nabídkách, a aktualizacích.
    CodedTrip

    Operated by CodedTrip LLC, USA.

    YouTube
    TikTok