Google releases Gemini 3.1 Pro: Benchmark performance, how to try it
Mashable
February 19, 2026
AI-Generated Deep Dive Summary
Google has launched its latest AI model, Gemini 3.1 Pro, showcasing significant advancements in reasoning capabilities. This updated version not only doubles the performance of its predecessor, Gemini 3 Pro, on the ARC-AGI-2 benchmark but also surpasses competitors like Claude and ChatGPT across various metrics. The release highlights Google's rapid pace in AI innovation, with the model now integrated into tools such as the Gemini App, API, and Notebook LM. Free users can access it through the app, while paid plans offer enhanced features.
Gemini 3.1 Pro excels in complex reasoning tasks, offering creative solutions to challenging problems. Its performance on benchmarks like Humanity's Last Exam and SWE-Bench Pro demonstrates a notable edge over models like GPT-5.2 and Claude Opus 4.6. Notably, it achieves a 77.1% score on ARC-AGI-2, significantly outpacing Gemini 3 Pro's 31.1%. This upgrade positions Gemini as a formidable competitor, potentially altering the AI landscape.
The model's availability varies by user tier: free users get basic access, while enterprise and paid plan users gain advanced features through platforms like Vertex AI and Gemini Enterprise. Developers can integrate it via APIs such as Gemini CLI and Android Studio, enabling broader adoption across industries.
Gemini 3.1 Pro's release comes amid intense competition in the AI sector. Its superior performance on key benchmarks could shift market dynamics, challenging OpenAI's dominance. The model's enhanced capabilities make it a valuable tool for professionals seeking advanced problem-solving and creative insights, solidifying Google's position as a major player in AI development.
This update underscores the
Verticals
tech
Originally published on Mashable on 2/19/2026