Gemini 3.1 Pro is available starting today in preview for developers, enterprises, and consumers and, according to Google, should deliver better performance in complex problem-solving and multi-step reasoning.
Gemini 3.1 Pro is available starting today in preview for developers, enterprises, and consumers. With this new version, Google is bringing enhanced core intelligence to its AI products, including the Gemini app, Vertex AI, and NotebookLM. The update builds on the Gemini 3 series and focuses on stronger reasoning capabilities. According to the company, the model should perform better in complex problem-solving and tasks requiring multiple reasoning steps.
High scores in reasoning
With Gemini 3.1 Pro, Google says it is taking a step forward in core reasoning. This is evident from results on various benchmarks. On ARC-AGI-2, a test that measures how well a model recognizes and solves new logical patterns, version 3.1 Pro achieves a verified score of 77.1%. That is more than double the score of Gemini 3 Pro.
The model is designed for scenarios where a simple answer is not enough. It focuses on applications such as visually explaining complex topics, synthesizing large amounts of data into a single overview, and supporting creative processes.
Additionally, the model can generate animated SVG files directly in code from a text prompt. Because these animations are built in pure code, they remain sharp at any scale and file sizes remain limited compared to traditional video files.
Available via API and enterprise platforms
Developers can access Gemini 3.1 Pro via a preview in the Gemini API within Google AI Studio, Gemini CLI, the Antigravity development platform, and Android Studio. For enterprises, the model is available via Vertex AI and Gemini Enterprise.
Consumers also get access. The rollout is taking place via the Gemini app, with higher usage limits for users with a Google AI Pro or Ultra subscription. NotebookLM offers the new version exclusively for the same subscription plans.
Google is releasing the update in preview to gather feedback and implement further improvements, including those around agentic workflows. General availability will follow later.