Gemini Robotics 1.5

Gemini Robotics 1.5: Bringing AI Agents into the Physical World

Gemini Robotics 1.5 is a cutting-edge AI model designed to bring advanced AI agents into the physical world. This model enables robots to perceive, plan, think, use tools, and act to solve complex, multi-step tasks. It is part of a family of models that aim to create more capable and versatile robots.

Benefits

• Advanced Thinking: Gemini Robotics 1.5 can reason about the physical world, create detailed multi-step plans, and use digital tools to complete missions. It achieves state-of-the-art performance across spatial understanding benchmarks.
• Transparency: The model shows its thinking process, making its decisions more transparent and helping robots assess and complete complex tasks more effectively.
• Learning Across Embodiments: Gemini Robotics 1.5 can transfer motions learned from one robot to another, accelerating the learning of new behaviors and making robots smarter and more useful.
• Safety: The model implements a holistic approach to safety, including high-level semantic reasoning and adherence to physical safety constraints.

Use Cases

Gemini Robotics 1.5 is designed to help robots complete complex, multi-step tasks in various environments. For example, it can sort objects into the correct compost, recycling, and trash bins by searching for relevant local recycling guidelines and performing the necessary steps. It can also sort laundry by color, demonstrating its ability to understand and execute multi-level thinking processes.

Additional Information

Gemini Robotics 1.5 is currently available to select partners. The model is part of a broader effort to advance intelligent, truly general-purpose robots. The development of Gemini Robotics 1.5 is guided by a commitment to responsible AI and robotics, ensuring that the models are developed in line with AI principles and safety guidelines.

The model has been evaluated on various benchmarks, showing state-of-the-art performance in spatial understanding and semantic safety. It marks an important milestone towards solving AGI in the physical world, moving beyond models that react to commands to create systems that can truly reason, plan, and generalize.

Gemini Robotics 1.5 is developed by a team of researchers and engineers, with contributions from various teams across Google and Google DeepMind. The model is part of a broader effort to advance intelligent, truly general-purpose robots, and the team is excited to continue this work with the broader research community.