OPEN LINK ↗
// 1h agoMODEL RELEASE
Gemini Robotics ER 2 Brings Cloud Reasoning
Gemini Robotics ER 2 is Google DeepMind’s embodied-reasoning model for robots, combining spatial understanding, multi-step planning, video progress tracking, tool use, and multi-robot coordination. Its Google Cloud availability lets developers connect these capabilities to custom robot APIs and lower-level vision-language-action models.
// ANALYSIS
ER 2’s biggest contribution is architectural: a reasoning layer supervises specialized robotic controllers, making complex tasks more adaptable across robot platforms. Preview status, cloud latency, and real-world reliability remain the key engineering hurdles.
- –Separating planning from motor execution makes task logic more portable across robot bodies.
- –Continuous video understanding and success tracking target long-horizon failure recovery.
- –Streaming, function calling, and custom tool use give developers practical integration points.
- –Google reports real-VLA task success improving from 48.6% with ER 1.6 to 60.0% with ER 2.
- –Physical deployments still need deterministic low-level safeguards alongside model-based reasoning.
// TAGS
gemini-robotics-er-2roboticsreasoningmultimodaltool-usecloudapi
DISCOVERED
1h ago
2026-10-27
PUBLISHED
1h ago
2026-10-27
RELEVANCE
9/ 10
AUTHOR
GoogleCloudTech