Google’s latest Gemini 3.1 Pro is capable of cracking puzzles it has never encountered before.
Key information to consider
- Google is showcasing Gemini 3.1 Pro, emphasizing advanced, multi-step problem-solving that confounds the majority of chatbots.
- The model achieved a score of 77.1% on the ARC-AGI-2 benchmark — over two times the reasoning capability of Gemini 3.0 Pro.
- It is accessible via the Gemini app, NotebookLM, Vertex AI, and the Gemini API.
Shortly after introducing Gemini 3.0, Google is now unveiling Gemini 3.1 Pro. This updated model is engineered to address complex, multi-step issues that most chatbots still struggle with.
This enhancement follows shortly after Google launched its Deep Think features, which broadened the platform’s capabilities in research, engineering, and science. Gemini 3.1 Pro powers these improvements, and Google is now integrating it into consumer applications, developer tools, and enterprise platforms simultaneously.
You can already utilize