Google unveiled Gemini 3.1 Pro, an upgraded AI model aimed at complex problem-solving, and began rolling it out in preview across consumer, developer and enterprise products. The company said the model marks a step-change in reasoning, posting a verified 77.1% on the ARC-AGI-2 benchmark—more than double the performance of Gemini 3 Pro—while introducing practical capabilities such as generating code-based animated SVGs. Developers can access the release via the Gemini API in Google AI Studio, CLI, Antigravity and Android Studio; enterprises through Vertex AI and Gemini Enterprise; and consumers via the Gemini app and NotebookLM, with higher limits for Pro and Ultra tiers. Google framed the preview as a validation phase ahead of general availability and flagged further work on agentic workflows. The launch follows last week’s release of Gemini 3 Deep Think, underscoring the company’s accelerated cadence in AI upgrades.
Related articles:
Vertex AI | Google Cloud
Android Studio
Toolformer: Language Models Can Teach Themselves to Use Tools
AI & Machine Learning | Google Cloud Blog
