Google DeepMind releases Gemini 4 Argon for robotics with 87% ERQA score
Original titleGemini 4 Argon for Robotics
AISummary
Google DeepMind releases Gemini 4 Argon, a model it describes as a general-purpose physical reasoning model for robotics. It scores 87% on the Embodied Reasoning (ERQA) benchmark, which the chart shows ahead of the other models it compares against.
The post shows prompts for robot control, real-to-sim simulation, physical understanding, and self-improvement, and gemini-4-argon is available in Google AI Studio, the Gemini API, and Google Antigravity.
AIWhy it matters
The post shows Gemini 4 Argon applied across the robotics stack, from control and simulation to data labeling, with prompts and benchmark results for each.
Source: Google DeepMind · deepmind.googlePublished · added here