Gemini Robotics 2.5 accelerates real-world general dexterity
Web page · https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence
Our most capable vision-language-action model enhances robot dexterity, reliability, and safety
First seen by Oriel on .
AI summary
Google DeepMind introduces Gemini Robotics 2.5, a vision-language-action model that improves robot dexterity, task performance, and safety across varied real-world tasks.
- Across 200 tasks, it achieved 30% to 100% higher success rates than previous baselines and completed tasks 20% to 30% faster.
- In a Mercedes-Benz production-line pilot, Apptronik’s Apollo 2 performed automotive kitting with a 96% success rate.
- Inference-time reasoning guidance increased safe task completion by 33% around obstacles and 43% around nearby people.
- The model supports multiple robot types, including Apollo 2 and Franka Duo, and is rolling out to robotics partners.
- Gemini 4 Argon accelerates development by generating simulated practice runs, building digital twins, and helping researchers investigate task failures.
#robotics #vision-language-action models #robot safety #simulation
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| Google DeepMind Blog | https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence/ | https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence/ Read discussion |