IllustrationNo official image on file. Showing a schematic category illustration.
Upgraded embodied-reasoning model with better visual/spatial understanding, instrument reading and safety awareness.
Key specs
| Specifications | Not stated |
|---|
Technical details
Gemini Robotics-ER 1.6 (April 2026) is DeepMind's reasoning-first VLM for robotics, specializing in visual and spatial understanding, task planning and success detection, and acting as a robot's high-level planner that calls tools such as Google Search, VLAs or user functions. DeepMind reports improvements over ER 1.5 and Gemini 3.0 Flash on pointing, counting and success detection, plus a new instrument-reading capability for gauges and sight glasses developed with Boston Dynamics. It was exposed via the Gemini API with a developer Colab, and was deprecated on August 31, 2026 in favor of ER 2.
| Architecture | Embodied-reasoning vision-language model (model code gemini-robotics-er-1.6-preview) with native tool calling (Google Search, VLAs, user-defined functions). |
|---|---|
| Inputs outputs | Inputs: text, images, video, audio (131,072-token input limit); output: text incl. points, boxes, plans (65,536-token output limit). |
| Embodiments supported | Robot-agnostic high-level reasoner; instrument-reading capability developed with Boston Dynamics. |
| License weights | Proprietary; API access only. |
| Hardware requirements | Cloud API (Gemini API / Google AI Studio). Deprecated on 2026-08-31 in favor of Gemini Robotics ER 2. |
| How to try | github.com/…/robotics-samples |
| Sources |
