Google announced the upgraded Gemini 3.1 Pro model—“suitable for complex tasks.”

Google launches Gemini 3.1 Pro – a model for complex scientific and technical tasks

*Google announced the release of an updated version of Gemini 3 Pro, naming it Gemini 3.1 Pro. The model is specifically designed for cases where a simple answer is insufficient and deep analysis is required.*

What’s new in 3.1 Pro?
ParameterDetailsVersionThe first Google model with an “x.1” number (previous updates were marked as “x.5”, e.g., Gemini 2.5 Pro).PurposeSolving high‑complexity problems in science, engineering, and applied research.Key focus“Basic Thinking” – the model’s ability to develop solutions independently without relying on templates.

Performance metrics
* In the ARC‑AGI‑2 benchmark (logical tasks not encountered before), Gemini 3.1 Pro scored 77.1 %, more than twice the result of Gemini 3 Pro.

Capabilities and use cases
FeatureDescriptionMulti‑step analysisVisual explanation of complex topics, synthesis of disparate data into a unified picture.Creative projectsGeneration of animated SVG graphics from text prompts (code, not video → scalability and small size).

Availability
* Public access – you can already try the model in the Gemini app and web version.

* Paid plans – Google AI Pro and Ultra users will receive higher limits and access via NotebookLM.

* Developers & enterprises – Gemini API (Google AI Studio), CLI, Antigravity, Android Studio; corporate clients can test the model in Vertex AI and Gemini Enterprise.

Current status
Gemini 3.1 Pro is in a pre‑release stage to gather feedback and refine features for ambitious agent workflows before making it fully available to the broader audience.