New Delhi: Google’s most advanced Gemini model yet is now out in the open, and early benchmark numbers are making noise. A new document detailing the Gemini 3 Pro model card and its performance on popular AI benchmarks has quietly surfaced online ahead of any formal press briefing. Published in November 2025, the model card confirms what many in the developer and research community had been expecting: Gemini 3 Pro is built to go big across tasks involving text, images, audio, video, and even code.

From AI-based reasoning to mathematical problem solving and code generation, the leaked benchmark chart shows the Gemini 3 Pro model outperforming not just its predecessor Gemini 2.5 Pro, but also top models like Claude Sonnet 4.5 and OpenAI’s GPT-5.1 in several core areas. The race between G

See Full Page