Google DeepMind · 2026-04-14 · major
Gemini Robotics-ER 1.6 — Instrument Reading Jumps from 23% to 93% with Agentic Vision
Google DeepMind's Gemini Robotics-ER 1.6 adds instrument reading (23%→93% with agentic vision), multi-view camera reasoning, and precision pointing. Available immediately via Gemini API. Purpose-built for robots that read physical gauges and interact with equipment.
Gemini's robotics reasoning model now reads physical gauges with 93% accuracy — up from 23% in the previous version.
Key specs
| Instrument reading (standard) | 86% |
|---|---|
| Instrument reading (agentic vision) | 93% |
| Vs er 1.5 baseline | 23% |
What is it?
Gemini Robotics-ER 1.6 is Google DeepMind's specialized spatial reasoning model for robotics, released April 14, 2026. It functions as the high-level reasoning layer for robots, capable of calling tools and processing multiple camera feeds simultaneously. The headline improvement is instrument reading: the ability to interpret physical analog gauges, pressure indicators, and sight glasses — a task that requires precise visual calibration absent from general-purpose vision models.
How does it work?
ER 1.6 uses agentic vision chains where the model makes multiple visual queries (zoom, point, compare readings) to triangulate precise readings from analog gauges, rather than attempting single-pass extraction. Multi-view reasoning processes several camera feeds in parallel for improved spatial understanding. The instrument reading benchmark shows: ER 1.5 at 23%, Gemini 3.0 Flash at 67%, ER 1.6 at 86%, and ER 1.6 with agentic vision at 93%. Available via Gemini API and Google AI Studio immediately.
Why does it matter?
Industrial robots need to read analog instruments — pressures, temperatures, fluid levels — that have no digital interface. Jumping from 23% to 93% instrument reading accuracy in one generation makes ER 1.6 viable for industrial inspection, manufacturing quality control, and facility maintenance tasks. The API availability means developers can integrate this into robotic pipelines today without running proprietary hardware.
Who is it for?
Robotics developers and industrial automation engineers building AI-powered inspection and control systems