Gemini 3.1 Pro: A smarter model for your most complex tasks
Key Points
- Advanced reasoning baseline (77.1% on ARC-AGI-2)
- Generates code-first assets like animated SVGs
- Preview rollout across API, Vertex AI, and Gemini app
Summary
Gemini 3.1 Pro is an upgraded core model focused on advanced reasoning and complex problem-solving. It achieves a verified 77.1% on ARC-AGI-2 (more than double the reasoning performance of Gemini 3 Pro) and is rolling out in preview across developer, enterprise, and consumer endpoints to power tasks that require multi-step synthesis, clear visual explanations, and design-aware code generation.
Key Points
- Improved core reasoning: 77.1% on ARC-AGI-2, significantly outperforming 3 Pro.
- Designed for complex tasks: data synthesis, multi-step explanations, agentic workflows, and system-level integration.
- Code-first outputs: generates website-ready animated SVGs and other compact, scalable assets.
- System synthesis examples: builds live telemetry dashboards (e.g., ISS orbit visualization) and bridges APIs to user-facing interfaces.
- Interactive and creative coding: can produce immersive 3D interactions and translate literary/creative themes into functional UI code.
- Preview release: being validated in preview before general availability.
Availability
- Developers (preview): Gemini API via AI Studio, Gemini CLI, Google Antigravity, Android Studio.
- Enterprises: Vertex AI and Gemini Enterprise.
- Consumers: Gemini app (higher limits for Google AI Pro and Ultra plans) and NotebookLM (Pro/Ultra).
Engineer notes: prefer 3.1 Pro when you need robust stepwise reasoning or system synthesis; validate outputs during the preview phase. Generative AI is experimental.