Google has officially launched Gemini 3, its newest and most powerful suite of AI models, succeeding Gemini 2.5 introduced earlier this year at Google I/O.
The company says Gemini 3 delivers major improvements across performance, multimodal understanding, and user interaction, marking Google’s biggest step forward in AI capabilities to date.
The rollout of Gemini 3 begins in AI Mode in Google Search, available to Google AI Pro and AI Ultra subscribers. It is also launching broadly in the Gemini app for all users. According to Google CEO Sundar Pichai, the new model is “much better at figuring out the context and intent behind your request, so you get what you need with less prompting”. This enhancement aims to make everyday AI interactions faster, smoother, and more intuitive.
Google describes Gemini 3 as “the best model in the world for multimodal understanding”, offering richer visual outputs, deeper interactivity, and improved comprehension across text, images, video, audio, and even code. The launch begins with Gemini 3 Pro, currently available in “preview”, with more advanced versions on the way.
Following Gemini 3 Pro, Google will introduce Gemini 3 Deep Think, which is now undergoing testing with safety evaluators. This model outperforms Gemini 3 Pro on most benchmarks but requires more processing time due to its deeper reasoning abilities. Once testing is complete, it will be released exclusively to Google AI Ultra subscribers.
Early benchmarks show that Gemini 3 Pro significantly outperforms Gemini 2.5 Pro, offering more detailed, thoughtful, and precise responses. Google says the model avoids unnecessary flattery and clichés, providing more direct and insightful answers tailored to what users truly need.
One of the key strengths of Gemini 3 is its ability to help users learn complex topics. Built from the ground up for multimodal synthesis, it can seamlessly integrate information across various formats, making explanations clearer and more comprehensive.
Google also claims that Gemini 3 is its most secure AI model yet. It has undergone the company’s most extensive safety evaluations, showing reduced susceptibility to prompt manipulation, stronger resistance to cyberattack misuse, and less sycophantic behavior.
With Gemini 3, Google aims to set a new industry standard in AI performance, safety, and multimodal intelligence, positioning the model as a major milestone in the future of AI-powered experiences.