Google has announced the release of its new flagship artificial intelligence model, Gemini 3.1 Pro, which is now available for early access. According to the company, this version is significantly smarter and more functional compared to its predecessors, including Gemini 3 Pro.
This is reported by Finway
Performance Improvements and New Features
Gemini 3.1 Pro surpasses the capabilities of the previous version in key benchmarks and offers expanded functionality for users. For instance, in the Humanity’s Last Exam test, which assesses knowledge depth across various fields, the new model scored 44.4%. In comparison, Gemini 3 Pro scored 37.5%, while GPT-5.2 achieved 34.5%.
The score for the ARC-AGI-2 benchmark was 77.1%, which is more than double the result of Gemini 3 Pro. The model also leads in 6 out of 10 major tests, demonstrating high efficiency and optimized token usage. Experts from Artificial Analysis noted a more effective token cost ratio compared to other flagship AIs.
“The new version is positioned as an improved Gemini 3 Pro. The optimization pertains to deeper reasoning, efficient token usage, and consistent responses.”
Technical Specifications and Innovations
Gemini 3.1 Pro supports working with text, audio, video, images, and PDF files. The model processes over 1 million input tokens and more than 65,000 output tokens. New features include the ability to generate scalable SVG animations from text descriptions, create interactive dashboards for complex systems, and build 3D simulations with hand motion tracking and real-time generative sound.
Gemini 3.1 Pro can also transform abstract literary ideas into functional visual interfaces. For example, based on the novel “Wuthering Heights,” the model not only retells the plot but also creates an interface that reflects the atmosphere of the work.
At the same time, the user community has noticed some shortcomings. Some experts pointed out that the model, like its predecessors, is prone to getting stuck in infinite loops when handling complex queries.
Starting February 19, 2026, Gemini 3.1 Pro will be available in early access for developers, enterprises, and regular users. An increased token limit is provided for subscribers of Google AI Pro and Ultra services.
