AI อะไรเนี่ย

Model

Gemini 3.1 Pro is here — Google's smarter model for complex tasks

Gemini 3.1 Pro is here — Google's smarter model for complex tasks

Google announced Gemini 3.1 Pro on Feb 19, 2026. It's the upgraded core intelligence behind last week's Gemini 3 Deep Think update — and it's rolling out everywhere: Gemini app, NotebookLM, Vertex AI, Gemini Enterprise, and the Gemini API in preview.

What it's for
3.1 Pro is built for tasks where a simple answer isn't enough: advanced reasoning, synthesizing data into one view, or explaining complex topics clearly. On ARC-AGI-2 (logic patterns the model hasn't seen before), 3.1 Pro hit 77.1% — more than double the reasoning performance of 3 Pro. So when you need the model to think rather than just reply, this is the one.

Why it matters

  • Code from words: It can generate website-ready, animated SVGs from a text prompt — crisp at any scale, tiny file size. No pixel exports, just code.
  • Real systems: In a demo it wired a public telemetry stream into a live aerospace dashboard and visualized the ISS orbit. That kind of "connect messy APIs → clean UX" is exactly where stronger reasoning pays off.
  • Creative + interactive: It can code a 3D starling murmuration where you manipulate the flock with hand-tracking and hear a generative score that reacts to the birds. Or turn a novel's mood (e.g. Wuthering Heights) into a modern portfolio design — not a summary, but a reasoned take on tone and interface.
    So: better for agentic workflows, dashboards, creative coding, and anything that needs a chain of reasoning, not a one-liner.

Where you get it

Google is shipping 3.1 Pro in preview to validate these updates and push further on agentic workflows before general availability. If you've been waiting for a model that can actually reason through hard tasks — not just answer — this is the one to try.

Try it: Fire up AI Studio and pick the 3.1 Pro preview model, or use it in the Gemini app if you're on Pro/Ultra.