
by Google
Google launches Gemini 3.1 Pro with 2x reasoning performance. ARC-AGI-2 score of 77.1%. First .1 increment in Gemini history.
[Google](https://startupintros.com/orgs/google) launched Gemini 3.1 Pro on February 19, 2026 — a model focused on "complex problem-solving" that achieves an ARC-AGI-2 score of 77.1%, more than double the reasoning performance of Gemini 3 Pro. The .1 increment is a first for Google, which previously used .5 for mid-cycle updates.
Gemini 3.1 Pro builds on the "upgraded core intelligence" from Gemini 3 Deep Think (Feb 12). It's rolling out across the Gemini app (all users), NotebookLM (Pro/Ultra subscribers), Google AI Studio, Vertex AI, Gemini CLI, and Android Studio.
Google is accelerating its model release cadence. The shift from .5 to .1 increments signals more frequent, iterative improvements rather than big-bang releases. For developers building on Gemini, this means faster access to reasoning upgrades — but also more frequent migration decisions.

Organizing the world’s information and making it accessible through search
private
1998