Google releases Gemini Robotics 2.5 with 96% success rate in production tasks

The model enhances vision, language, and action control for real-world operation. It outperforms prior models across 200 diverse tasks including food packaging and tying ribbons. Inference-time safety guidance improves safe operation in obstacle- or human-present environments to 33% and 43%, respectively. The model supports deployment across multiple robot platforms with consistent performance.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.