Google Introduces Gemini Robotics 1.5 for Safer, Smarter Robots

News

1. Intel Courts Apple, TSMC After Nvidia’s $5 Billion Lifeline

2. Google Introduces Gemini Robotics 1.5 for Safer, Smarter Robots

3. Alibaba Takes Aim at Rivals With Qwen3 Max at 1 Trillion Parameters

4. OpenAI, Oracle, and SoftBank Partner for 5 Data Centers Under Stargate

5. Towards Superintelligence: Nvidia Puts $100B in OpenAI

6. India’s Next GCC Hubs May Rise Outside Big Cities

Google is taking a bold step into the future of robotics with the launch of two new AI models, Gemini Robotics 1.5 and Gemini Robotics-ER 1.5, that promise to bring intelligent, general-purpose robots closer to reality. These models, powered by the company’s flagship Gemini AI family, are designed to help robots not just follow commands, but think, plan, and act in the physical world.

Earlier this year, Google made headway in merging Gemini’s multimodal understanding with robotics. Now, with these upgrades, the company says it is unlocking “agentic experiences” that allow robots to actively reason about their environment and complete multi-step tasks. 

At the core of this innovation are two specialized models that work hand in hand. Gemini Robotics-ER 1.5, the “thinking” model, excels at embodied reasoning, acting like a high-level brain for robots. It can break down a mission into detailed steps, using digital tools like Google Search, and explain the physical world. With state-of-the-art performance in spatial understanding benchmarks, this model is adept at orchestrating complex plans.

Its counterpart, Gemini Robotics 1.5, focuses on execution. Described as the company’s most capable vision-language-action model yet, it translates instructions into real-world actions while explaining its reasoning process in natural language. Unlike traditional systems that move straight from command to motion, this model “thinks before acting,” making robots more transparent and reliable.

For instance, a robot sorting laundry doesn’t just follow orders. It reasons that whites go into one bin, colors into another, and then plans each movement, such as how to grip a sweater, before carrying it out.

A major breakthrough lies in the models’ ability to transfer skills across different robots. Training tasks on one robot, such as the ALOHA 2, can now be replicated on entirely different machines, including humanoid or bi-arm robots, without retraining. This cross-embodiment learning accelerates skill development, making robots more adaptable and useful in diverse environments.

With robots stepping into human-centric spaces, safety is a top priority. Google says it is proactively building safeguards through its Responsibility & Safety Council and its upgraded ASIMOV benchmark, which measures semantic safety and physical constraints. Gemini Robotics-ER 1.5 reportedly achieves state-of-the-art performance on these safety tests, ensuring robots can think about safety before acting, avoid collisions, and maintain respectful interactions with humans.

Experts see this as more than just an upgrade. By enabling robots to reason, plan, and act with greater autonomy, Google is edging closer to embodied artificial general intelligence (AGI), machines that can operate with human-like intelligence in the physical world.

While Gemini Robotics-ER 1.5 is available today to developers via the Gemini API in Google AI Studio, Gemini Robotics 1.5 will be rolled out to select partners first. Google hopes that the broader robotics community will build on these models to create more capable, versatile, and safe robots for everyday life.