Google announced on Wednesday that Gemini can now generate interactive 3D models and simulations directly within its responses, a capability the company says moves AI assistants beyond flat text and static images [1].

The feature allows users to ask Gemini questions about spatial concepts — molecular structures, architectural layouts, mechanical systems, geographic terrain — and receive manipulable 3D objects they can rotate, zoom, and explore within the chat interface. Google demonstrated examples including a rotating model of a protein structure, a cross-section of a jet engine, and a topographic simulation of watershed drainage.

"This is about making AI responses match how people actually think about the physical world," said Sissie Hsiao, Google's vice president for Gemini experiences. The 3D generation runs on a new rendering pipeline built into the Gemini infrastructure, producing lightweight models optimized for browser and mobile display.

The practical applications are most obvious in education and professional contexts. A medical student asking about cardiac anatomy can now manipulate a 3D heart rather than reading a paragraph. An engineer exploring stress distributions on a beam can see the simulation rather than interpreting a table of numbers.

Google is positioning the feature as a differentiator in the increasingly crowded AI assistant market. OpenAI's ChatGPT and Anthropic's Claude still respond primarily in text and images. While those platforms offer code execution and data visualization, none currently produce interactive 3D objects.

The capability is rolling out to Gemini Advanced subscribers this week, with broader availability planned over the coming months.

-- DAVID CHEN, Beijing