Google has introduced Gemini 3, its latest artificial intelligence model, on 18 November 2025. Developed by Google DeepMind, the system features expanded reasoning capabilities, higher accuracy and enhanced multimodal processing, and is now available across Google Search, the Gemini app, Google AI Studio, Vertex AI and the new developer platform, Google Antigravity.
Gemini 3 is described by Google as its most advanced AI system to date, designed to process text, images, audio, video and code within a single workflow. The model supports a one million-token context window, enabling it to review lengthy documents, research material and videos at once. According to the company, the system provides faster responses and improved handling of long-form inputs compared with previous versions.
A key addition is Deep Think mode, an optional setting intended for scientific, analytical and complex coding tasks. The feature will be accessible to Google AI Ultra subscribers once fully rolled out. Google says the new model is also intended to offer a more structured experience for developers, allowing them to create applications, websites, games and full-stack tools using natural-language prompts.
As part of the broader update, Google has also introduced Antigravity, a developer environment focused on agent-based coding. The platform is built to support planning, writing and testing code, conducting browser-based simulations and automating software development workflows.
Compared with Gemini 2.5, the new system shows improvements in multimodal accuracy, long-horizon planning and reasoning benchmarks, according to Google’s internal assessments. The company says Gemini 3 provides more stable one million-token support across major apps, expanded agent tools and improvements in video, image and audio understanding.
The shift toward agent-based development marks a move from single-prompt interaction to goal-oriented workflows, where AI systems handle end-to-end execution with limited human intervention. Google highlights that Gemini 3 is positioned to serve both consumer and enterprise users across education, research, development and productivity-focused tasks. |
