Web Desk — November 19, 2025
San Francisco: Google has unveiled Gemini 3, its most advanced AI model to date, billed as the company’s smartest and most factually accurate system ever. Launched just seven months after Gemini 2.5, the new model is immediately available to all users via the Gemini app, AI search interface, and developer tools, marking a swift and widespread rollout.
Telsi Doshi, Senior Director and Head of Product at Google DeepMind, described Gemini 3 as a pivotal step in Google’s mission to make information “comprehensive and usable.” She highlighted its shift beyond text-only responses, offering users a richer, more interactive view of the world—processing text, images, and audio simultaneously to create dynamic outputs like interactive flashcards from video lectures.
At its core, Gemini 3 Pro is a “native multimodal” powerhouse, capable of handling complex queries with unprecedented reasoning. It shattered records on key benchmarks: scoring 37.4 on Humanity’s Last Exam (surpassing GPT-5 Pro’s 31.64) and leading LMSYS Arena for user satisfaction. Google claims 72% accuracy on standard tests, emphasizing its edge in coding, creative tasks, and agentic functions—like generating custom interfaces for prompts such as “build a Van Gogh gallery with contextual bios.”
The upgrades ripple across Google’s ecosystem: enhanced AI Overviews in Search for premium subscribers, a new Canvas tool in the Gemini app for advanced projects, and upcoming “DeepThink” mode for Ultra users. A dedicated coding app accompanies the launch, streamlining developer workflows.
This aggressive debut positions Google firmly in the AI arms race against rivals like OpenAI and Anthropic, with Sundar Pichai teasing more Gemini 3 variants soon. Early feedback is pouring in—users are already experimenting with its “generative interfaces” for everything from art curation to lesson planning. As AI evolves, Gemini 3 promises to make the tech feel less like a chatbot and more like a creative companion.
