Google DeepMind ships Gemini 3.1 Pro to harden core intelligence across consumer and developer surfaces. Google DeepMind positions the release as a direct lift in logic and synthesis capability, rolling into the Gemini app, NotebookLM, Vertex AI and agentic tools. Code and data workflows gain a sharper baseline that treats complexity as standard rather than edge case.

Progress follows November’s Gemini 3 Pro debut, with ARC-AGI-2 results showing 77.1% verified performance, more than double the prior model. Tasks span live telemetry bridging, animated SVG generation and immersive 3D scenes, turning abstract prompts into deployable artifacts. Preview access validates updates while tightening reliability for high-stakes reasoning.

Developers and enterprises obtain early entry through Gemini API, Android Studio, CLI and Vertex AI, while Pro and Ultra subscribers unlock higher limits in consumer apps. Rigorous benchmarks now back practical synthesis of APIs, visuals and narrative tone into cohesive interfaces. The model exits preview only after proving stability under real workloads.