# Gemini Robotics ER 2 reaches public preview on Google Cloud

published: 2026-10-27T17:05:27.383Z
topic: infrastructure
source: https://cloud.google.com/blog/products/ai-machine-learning/gemini-robotics-embodied-reasoning-2-is-now-available-on-google-cloud/
url: https://aiagentnews.io/stories/gemini-robotics-er-2-reaches-public-preview-on-google-cloud

Google Cloud has made Gemini Robotics Embodied Reasoning 2 available in public preview on the Gemini platform. The model lets developers register robot control interfaces as tools and deploy the resulting physical agents with Google Cloud's enterprise security and lifecycle tooling.

Google Cloud has made Gemini Robotics Embodied Reasoning 2, a specialized Gemini model that acts as a high-level planner for robots, available in public preview on the Gemini platform. The model is aimed at robots of many kinds, from stationary arms to full humanoids, and combines spatial and temporal video understanding with long-horizon planning. Developers can access it through Model Garden.

Developers declare low-level control interfaces, such as Vision-Language-Action models, navigation APIs or custom functions, as tools, and stream video, audio or text into the model. Google reports 91.3% accuracy when pinpointing the exact video frame of a critical event, with a 0.96-second mean absolute frame distance, running at 4x the execution speed of many larger model categories.

For deployment, the Gemini platform provides enterprise IAM, security controls and built-in function calling. Latency-sensitive tasks such as SLAM and path planning can run on factory floors through Google Distributed Cloud, and fleet telemetry can be streamed into BigQuery. Simulation is supported in MuJoCo-Warp and NVIDIA Isaac Sim.

## Why it matters if you build agents

Teams building robot agents can call a reasoning model for task verification, self-correction and multi-robot coordination without building that layer themselves. Because the model is in public preview, the documentation and pricing should be checked before it is used in production workloads.