Google DeepMind launches Gemini Robotics-ER 1.5 for smarter robot reasoning

Google DeepMind has released Gemini Robotics-ER 1.5, a new embodied reasoning model designed specifically for robotic applications that can perform spatial reasoning, multi-step planning, and execution in physical environments. The model represents a significant shift toward separating reasoning from execution in robotics AI, potentially making it easier to adapt advanced AI capabilities across different hardware platforms and accelerating the development of general-purpose robots.

What you should know: Gemini Robotics-ER 1.5 is now available in preview through Google AI Studio and the Gemini API, offering developers a new approach to robotic intelligence.

• The model can output precise 2D points grounded in object size, weight, and affordances, supporting commands such as “point at any object you can pick up.”
• Developers can adjust a “thinking budget” to balance response latency with reasoning accuracy, giving them control over performance trade-offs.
• Built-in safeguards prevent unsafe or physically infeasible plans by checking payload limits and workspace constraints.

How it works: The system operates as a dual-model architecture that separates high-level reasoning from physical execution.

• While Gemini Robotics-ER 1.5 doesn’t directly control robot actuators, it can call external tools such as vision-language-action (VLA) models or user-defined functions to execute commands.
• The design allows robots of different configurations to share higher-level reasoning abilities through a single adaptable software stack.
• This separation between reasoning and execution could make the system more flexible across different hardware platforms compared to previous direct perception-to-action mapping approaches.

Performance metrics: DeepMind, Google’s AI research division, reports that the system achieved state-of-the-art performance across 15 robotics benchmarks and demonstrated compatibility with platforms ranging from dual-arm lab robots to humanoids.

Who else is involved: Partners like Apptronik, a humanoid robotics company, and more than 60 testers are currently working with the system, indicating significant industry interest in the technology.

Competitive landscape: Gemini Robotics-ER differs from other large models applied to robotics, such as the Nvidia VLA, by emphasizing controllable reasoning depth and safety mechanisms rather than focusing solely on direct perception-to-action mapping.

What they’re saying: Industry experts see transformational potential in the general-purpose approach.

• “This general purpose approach will be transformational for robotics. Obviously the big robotic companies would partner with Google but even small ones could just license the AI from Google and build their own robots for solving niche problems,” said Sonia Sarao, an AI consultant.
• Brian Orlando, a CPA, described it as a pivotal moment: “Wild. Robots reasoning, planning, transferring skills. Feels like the real inflection point. Laundry today, general-purpose tomorrow.”

Mixed reactions: While robotics engineers highlighted the tunable reasoning budget as useful for balancing accuracy and deployment speed, some raised concerns that separating reasoning from actuation could add latency to robot operations.