Google has revealed the next generation of its AI technology with the launch of the Gemini 3 model and a new agent-driven development tool called Antigravity. Both were announced on November 18, highlighting Google’s push toward more autonomous, agent-based workflows for developers.
Gemini 3 is an upgraded version of the previous Gemini models, offering stronger reasoning skills and better handling of visual information. Google says the model is designed to follow instructions more reliably, call functions accurately, and support the creation of advanced software agents. Agentic features in Gemini 3 Pro are already being built into Google AI Studio, the Gemini command-line tools, Android Studio, and a growing number of third-party developer platforms.
The company claims Gemini 3 Pro outperforms Gemini 2.5 Pro in coding tasks and is especially good at agentic workflows and solving complex problems without example prompts. Developers can use it to quickly turn ideas into working applications thanks to its multimodal abilities. During the preview period, Gemini 3 is priced at two dollars per million input tokens and twelve dollars per million output tokens for prompts up to 200,000 tokens. It is accessible through the Gemini API, Google AI Studio, and Vertex AI for enterprise customers.
Google also introduced Antigravity, a new development environment where the AI agent takes the lead instead of being a small feature inside an IDE. Antigravity can control the browser, the code editor, and the terminal to complete a wide range of programming tasks. Google describes it as the next step in evolving IDEs toward an agent-first approach, with asynchronous interactions and the ability to manage entire workflows automatically. Antigravity is available in a public preview for Windows, macOS, and Linux.
With Antigravity, developers act more like project architects while AI handles the execution. The agent can work through full coding tasks, research problems, and provide progress updates along the way. Google says the system improves over time with feedback, thanks to built-in memory that helps it learn and adapt within each project.
