Gemini 4 Argon: Google's physical reasoning model for robotics

Google released Gemini 4 Argon, a general-purpose physical reasoning model aimed at robotics. A Google Developers Blog post dated Oct. 13, 2026, says it covers physical understanding, code generation and visual-semantic reasoning, and it is available to developers.

Google says Argon marks a step change in those areas and can be applied across the whole robotics lifecycle. It scores 87% on the Embodied Reasoning (ERQA) benchmark, which tests physical and spatial problems in 3D environments, using pure reasoning. Google says that outperforms every other frontier model.

The post groups example prompts into four areas: robot control, Real2Sim, physical understanding, and research and self-improvement. In one demo, Argon controls a Boston Dynamics Spot to step on a foot-pedal trash bin, practicing first in a MuJoCo digital twin. In another, it steers robots in the Gemini Omni World Model to arrange Scrabble tiles into a row.

Coverage from other outlets describes Gemini 4 Argon as rolling out more broadly, including to the Gemini app, AI Mode in Search and Google Antigravity, with AI Pro and AI Ultra users named for the app.