Google has released the preview version of Gemini 3.1 Pro via the Gemini App, LM Notebook, Vertex AI, and Gemini API. Google highlights the 77.1 percent verified score on the ARC-AGI-2 test and calls it a step forward in fundamental reasoning.
Google says it’s launching Gemini 3.1 Pro, describing it as the “updated core intelligence” behind the recent “deep thinking” update in Gemini 3, and positioning it as a smarter baseline for “tasks where a simple answer isn’t enough.”
Gemini 3.1 Pro is rolling out starting today across multiple products: Developers (preview): Gemini API via Google AI Studio, Gemini CLI, Google Anti-Gravity, and Android Studio. Enterprises: Vertex for AI and Gemini for Enterprise. Consumers: Gemini App and LM Notebook.
Google highlights its verified ARC-AGI-2 score of 77.1 percent. Google’s main performance claim for the Gemini 3.1 Pro is its advances in ARC-AGI-2, which Google describes as a benchmark that evaluates a model’s ability to solve entirely new logic patterns. Google says the Gemini 3.1 Pro achieved a verified score of 77.1 percent on this test, and calls that “more than double” the inference performance of the Gemini 3 Pro.
For reference, Google DeepMind’s official benchmarks table for the Gemini 3 Pro lists a 31.1 percent “ARC-AGI-2” (ARC Verified Award) score for the Gemini 3 Pro.
What Google says Gemini 3.1 Pro is designed to do Google frames Gemini 3.1 Pro as a model that aims to apply advanced reasoning to real workflows, such as: creating website-ready SVG animations from a text prompt, building a live dashboard that visualizes the ISS telemetry stream, producing code-driven interactive experiences (Google example: 3D simulation of a flock of starlings).
Availability details and plan limits Google says Gemini 3.1 Pro is launching in a preview mode where it validates updates before making it generally available “soon.” On the consumer side, Google adds that Gemini 3.1 Pro in the Gemini app is rolling out with higher limits to users on Google’s AI Pro and Ultra plans, while access to the LM Notebook is listed as exclusive to Pro and Ultra users.
What we can’t yet confirm from Google’s post alone Google’s post focuses on the most notable rollouts and standards, but it doesn’t include a full spec sheet with details like pricing, availability by country, or exact API model names. For these details, Google directs developers and companies to the official documentation for the Gemini and Vertex AI APIs.
