Google has launched Gemini Robotics ER 2, a new embodied reasoning model designed to help robots think fast and plan multi-step tasks in real-time. The update introduces significant upgrades over the previous ER 1.6 version, including continuous video-based progress tracking, precision moment-finding, and native multi-robot collaboration capabilities.
- Progress classification achieves 57.4% accuracy across five levels to help robots adjust actions on the fly.
- Moment-finding reaches 91.3% accuracy with a 0.96s mean absolute distance, enabling precise task switching at 4x the execution speed of previous models.
- Multi-robot collaboration allows diverse machines to communicate via shared semantic understanding to complete complex workflows.
- Spatial intelligence improvements include success/failure detection on raw video and generalized instrument reading across 10 types.
- Safety benchmarks show gains in instruction following and human proximity detection, allowing robots to halt when people are nearby.
The model is now publicly available to developers via the Gemini API and Google AI Studio, enabling more useful physical AI tasks through fluid orchestration without stop-and-think pauses.