2 months ago
48
ARTICLE AD BOX
Google has unveiled Gemini 3.1 Pro, a almighty AI upgrade emphasizing enhanced reasoning for analyzable tasks. This caller exemplary excels astatine synthesizing data, generating animated SVGs from text, and tackling intricate method problems. It importantly outperforms erstwhile versions connected cardinal benchmarks, promising a leap successful AI capabilities for some consumers and developers.
Google has launched Gemini 3.1 Pro, a important update to its flagship AI exemplary focused connected sharper reasoning and amended handling of complex, multi-step tasks. The exemplary is rolling retired contiguous crossed the Gemini app, NotebookLM, and developer platforms including Google AI Studio, Vertex AI, and Antigravity.
Google Gemini 3.1 Pro features: What the upgrade really brings
This is Google's first-ever .1 increment—previous cycles utilized .5 arsenic the mid-year refresh. The naming signals a focused quality upgrade alternatively than a wide diagnostic expansion. At its core, 3.1 Pro brings the precocious reasoning motor introduced successful Gemini 3 Deep Think past week to a overmuch wider audience.Google describes the exemplary arsenic built for situations wherever "a elemental reply isn't enough." In practice, that means handling things similar synthesising ample datasets into a azygous view, generating animated SVGs straight from a substance prompt, oregon moving done layered method and technological problems.
The animated SVG capableness is simply a neat example—outputs are built successful axenic codification alternatively than pixels, meaning they enactment crisp astatine immoderate standard with tiny record sizes compared to accepted video.
Google Gemini 3.1 Pro vs OpenAI GPT and Anthropic Claude: What the numbers say
The header fig is simply a 77.1% people connected ARC-AGI-2, a benchmark that tests abstract reasoning connected wholly caller logic patterns. That's much than treble Gemini 3 Pro's score—a meaningful jump. Elsewhere, 3.1 Pro scores 94.3% connected GPQA Diamond (scientific knowledge), 80.6% connected SWE-Bench Verified (agentic coding), and 85.9% connected BrowseComp (agentic search).
On LiveCodeBench Pro's competitory coding Elo, it sits astatine 2887—notably up of rivals.
Google Gemini 3.1 Pro: How to entree it close now
Gemini 3.1 Pro is launching successful preview portion Google validates performance, peculiarly astir agentic workflows. Consumers tin entree it done the Gemini app and NotebookLM, with higher usage limits for Google AI Pro and Ultra subscribers. Developers get entree via the Gemini API successful Google AI Studio, Gemini CLI, Antigravity, Vertex AI, Gemini Enterprise, and Android Studio. General availability is expected soon.
