Google establishes firm lead with new Gemini model, pulls ahead of competition
Google has just leapfrogged Anthropic and OpenAI with a new Gemini model, demonstrating a massive jump in capabilities across all benchmarks.
Google has officially released Gemini 3.1 Pro, its most advanced AI model, and says that it demonstrates more than double the reasoning performance in some benchmarks, compared to the previous model 3 Pro.
“We are shipping 3.1 Pro across our consumer and developer products to bring this progress in intelligence to your everyday applications,” Google announced.
Independent tests confirm a massive jump in Gemini’s capabilities.
The 3.1 Pro scores an average of 57 (out of 100) across ten different “intelligence” valuations, compared to 48 points achieved by the 3 Pro, as measured by Artificial Analysis.
Gemini now firmly surpasses the previous leader, the recently introduced Claude Opus 4.6 by Anthropic, which scores 53 points. Meanwhile, the most advanced OpenAI model, GPT-5.2, completes benchmarks with 51% accuracy.
Google itself says that the most impressive performance jump is on ARC-AGI-2, a benchmark that evaluates the model’s ability to solve entirely new logic patterns.
“3.1 Pro achieved a verified score of 77.1%. This is more than double the reasoning performance of 3 Pro,” Google boasts.
Some benchmarks that previously seemed challenging and required scientific expertise across many fields are now nearly saturated. For example, Gemini achieved over 94% accuracy in GPQA Diamond, whereas human PhD experts achieved only 65% accuracy.
Anthropic’s Opus 4.6 remains the best coding and tool-use agent, but Gemini 3.1 Pro has nearly closed the gap, according to the slide provided by Google.
The tech giant says it designed the new Gemini for the hardest challenges, advanced reasoning, and tasks where a simple answer isn’t enough. The blog post demonstrates improved capabilities to generate “website-ready” SVG animation, complex dashboards, interactive designs, and other interfaces.
“What stands out about Gemini 3.1 Pro is how deeply it understands the vibe behind a user’s prompt. In Hostinger Horizons – our vibe-coding product for non-developers – we see that intelligence translates into code that reflects direction, style, and product intent, not just syntax. The improvement in reasoning is immediately noticeable,” said Dainius Kavoliunas, Head of Product at Hostinger Horizons.
The 3.1 Pro is currently in preview and rolling out to users with higher limits, using the Google AI Pro and Ultra plans. Developers can access the model in preview via the Gemini API in Google AI Studio, Android Studio, Google Antigravity, and Gemini CLI.
Unlock exclusive Cybernews content on YouTube.
