Google releases Gemini Robotics 2.5 and a safety benchmark
What happened
On November 10, 2026, Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model it says improves how fast robots learn skills, plus dexterity and safety, with gradual access for robotics partners. That evening Ars Technica reported that DeepMind also published an open safety benchmark, ASIMOV-jailbreak. The platform has Gemini Embodied Reasoning plan tasks from instructions and live video, then Gemini Robotics generates the actions. Google says the new version performs better on high-dexterity tasks, with evaluation covering more than 200 activities. On November 11, The Decoder reported further that Google says the model reached a 96% success rate in a pilot sorting metal parts on a Mercedes production line.
Written by AI from the coverage · updated yesterday
Coverage
Follow the reports to see the story from different sides.
- The DecoderGoogle DeepMind's Gemini Robotics 2.5 scores 96 percent sorting metal parts in a Mercedes-Benz assembly pilot
Google DeepMind 推出机器人控制模型 Gemini Robotics 2.5，并称其在奔驰实际生产线的金属零件分拣试点中取得96%的成功率。
- Ars Technica · AIDeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
Google DeepMind 发布 Gemini Robotics 2.5，并公布开源安全基准 ASIMOV-jailbreak，为机器人与人类协作做准备。平台由 Gemini Embodied Reasoning 根据指令和实时视频规划任务，再由 Gemini Robotics 生成机器人动作；Google 称新版在高灵巧度任务上表现更好，评估覆盖超过 200 项活动。
- Google DeepMindPickGemini Robotics 2.5 boosts general dexterity in real-world robot tasks
Google DeepMind released Gemini Robotics 2.5, a vision-language-action model it says improves how fast robots learn skills, along with dexterity and safety. The model is being opened gradually to robotics partners.
Heat over time
Heat now 8·Comparable peak 24（Nov 11 02:00）·Comparable change over 24 hours -61%
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.