Google announced the preview release of Gemini 3.1 Pro, describing it as an upgrade to the core intelligence behind its recent Gemini 3 Deep Think update and positioning the model for more complex reasoning tasks across consumer and enterprise products.
The new model is being made available in preview through the Gemini app and NotebookLM for consumers, as well as through Vertex AI and Gemini Enterprise for business users. Developers can access Gemini 3.1 Pro via the Gemini API in Google AI Studio, Gemini CLI, Google Antigravity and Android Studio. Google said the model is intended for scenarios “where a simple answer isn’t enough,” framing it as a stronger baseline for demanding workflows.
See also: Gemini 3 Deep Think Upgrade Targets Frontier Scientific Problems
A key performance claim centers on the ARC-AGI-2 benchmark, which measures a system’s ability to solve novel logic patterns. Google said Gemini 3.1 Pro achieved a verified score of 77.1% on ARC-AGI-2, more than double the 31.1% score listed for Gemini 3 Pro in benchmark results published by Google DeepMind.
According to Google, Gemini 3.1 Pro is designed to apply advanced reasoning in practical settings, including generating website-ready animated SVG graphics from text prompts, building live dashboards that visualize International Space Station telemetry streams, and creating interactive, code-driven experiences such as a 3D starling murmuration simulation.
See also: Google Introduces Custom Music Creation to Gemini Users
The company said the model is launching in preview as it continues validating updates before a broader general release. Higher usage limits in the Gemini app are being rolled out to subscribers on Google AI Pro and Ultra plans, while NotebookLM access to Gemini 3.1 Pro is limited to those tiers. Google’s announcement did not include detailed pricing, country availability or specific API model names, directing developers and enterprises to official documentation for further specifications.
Source: Gemini 3.1 Pro, Google Deepmind, Notebook LM, Gemini API