Gemini Robotics ER 2

Gemini Robotics ER 2 is Google DeepMind's most capable embodied reasoning (ER) model, launched on July 30, 2026. It is a vision language model (VLM) that acts as a robot's high-level brain: it communicates with humans, understands the physical world, and plans multi-step tasks lasting several minutes, then hands off motor execution to any lower-level vision-language-action (VLA) model. It succeeds Gemini Robotics ER 1.6 (and the earlier ER 1.5 tracked in EmbodiedHub). Key capabilities: continuous video understanding and progress tracking (progress classification: 57.4% accuracy across five progress levels; moment finding: 91.3% accuracy with 0.96s mean absolute distance at 4x execution speed), enabling robots to track task progress, self-correct, and retry failed steps without restarting a workflow. It can natively call tools like Google Search or any user-defined function, and it integrates with the Gemini Live API (bidirectional streaming) for latency-sensitive orchestration. Multi-robot collaboration: Gemini Robotics ER 2 enables different types of robots to communicate via shared semantic understanding and complete complex workflows together, demonstrated with Apptronik's Apollo 2 and Franka F3 Duo, and with Boston Dynamics Spot (orchestrating Spot APIs to fetch objects on natural language commands). Availability: publicly available to developers via the Gemini API and Google AI Studio, in private preview on Gemini Enterprise Agent Platform, with open examples in the google-gemini/robotics-samples GitHub repository. Safety: Google DeepMind reports it as its safest robotics model to date on Safety Instruction Following and Human Proximity benchmarks, evaluated in part by the new ASIMOV-Agentic benchmark for agentic safety orchestration.