Sun Dec 07 01:40:00 UTC 2025: FOR IMMEDIATE RELEASE
Google Unveils Gemini 3 Pro: A Generational Leap in Multimodal AI
MOUNTAIN VIEW, CA – December 5, 2025 – Google today announced the release of Gemini 3 Pro, its most advanced multimodal AI model to date. Gemini 3 Pro delivers state-of-the-art performance across a range of complex tasks, including document processing, spatial reasoning, screen understanding, and video analysis. Google touts it as a “generational leap” from basic recognition to true visual and spatial reasoning capabilities.
The model boasts significant improvements on key vision benchmarks like MMMU Pro and Video MMMU. It excels in deciphering messy, unstructured documents, performing highly accurate Optical Character Recognition (OCR), and understanding complex visual relationships. Gemini 3 Pro can even “derender” visual documents, reverse-engineering them into structured code like HTML, LaTeX, or Markdown.
In detailed testing, Gemini 3 Pro demonstrated superior reasoning skills in analyzing complex reports, outperforming human baselines on benchmarks like CharXiv Reasoning. It can accurately extract information from tables and charts, identify causal relationships, and perform numerical comparisons to answer complex questions.
Beyond documents, Gemini 3 Pro showcases impressive spatial understanding. Its ability to comprehend desktop and mobile OS screens makes it a robust tool for automating repetitive tasks and enhancing user interface understanding for quality assurance, user onboarding, and UX analytics.
Video understanding is another area where Gemini 3 Pro shines. The model processes video at 10 frames per second, unlocking deep insights into complex activities and allowing it to trace cause-and-effect relationships. Gemini 3 Pro can even translate knowledge from long-form video content into functioning apps or structured code.
Google envisions numerous applications for Gemini 3 Pro across various fields. In education, it can tackle complex diagrams and reasoning problems in math and science. It has potential for medical and biomedical image understanding, achieving state-of-the-art results on benchmarks like MedXpertQA-MM and VQA-RAD. Professionals in finance and law can leverage its enhanced document understanding to streamline complex workflows.
Developers can now fine-tune the model’s performance and cost via a new media_resolution parameter, allowing them to balance fidelity against resource consumption.
Gemini 3 Pro is available now through developer documentation and Google AI Studio. Google emphasizes that the model is not intended for clinical diagnosis or patient care and should not be a substitute for professional medical advice.
About Google
Google’s mission is to organize the world’s information and make it universally accessible and useful. Through products and platforms like Search, Maps, Gmail, Android, Google Play, Chrome and YouTube, Google plays a meaningful role in the daily lives of billions of people and has become one of the best-known companies in the world. Google is a subsidiary of Alphabet Inc.
First Piper 