Google Launches Gemini 3 Pro AI Model with Advanced Capabilities
- Google’s Gemini 3 Pro AI model can process up to one million tokens of context, equivalent to about ten full-length novels.
- Gemini 3 Pro scored significantly higher than its predecessor on benchmarks, achieving a score of 37.5% on Humanity’s Last Exam compared to Gemini 2.5 Pro’s score of 21.6%.
- The model utilizes a sparse mixture-of-experts architecture, activating only specialized subnetworks for each query to reduce computational costs.
- Google offers free access to Gemini 3 Pro through Google AI Studio, but it is not yet available via the Gemini app.
Google has released its most advanced AI model, Gemini 3 Pro, which processes text, images, audio, and video simultaneously while handling extensive amounts of data efficiently through specialized subnetworks.
With improved benchmark scores over its predecessor and free public access through Google AI Studio, Gemini positions itself as a strong contender against models like OpenAI’s GPT-5.1 in the competitive AI landscape.(Source)
