Daily Humanoid Robotics and AI Wrap: October 11, 2025
Figure AI Unveils Figure 03, Targeting Home Use and Mass Production with ‘Helix’ AI
Figure AI, a leading humanoid robotics startup, has introduced its third-generation robot, the Figure 03, marking a significant step toward developing a truly general-purpose machine for both industrial and domestic environments. The new model represents a comprehensive, ground-up redesign of both the hardware and software systems, emphasizing its suitability for mass production and household integration.
The company has centered the Figure 03’s architecture around its proprietary vision-language-action (VLA) AI, dubbed “Helix,” which is designed to enable real-time reasoning and control in the physical world. The robot’s hardware upgrades are specifically tailored to bring the Helix AI to life. A key focus is on enabling the robot to master complex, human-like tasks without relying on predetermined scripts.
In a demonstration video, the 5-foot-6-inch, 132-pound robot showcased a range of domestic capabilities, including folding clothes, watering plants, and loading a dishwasher. Company CEO Brett Adcock affirmed that the demonstrations were fully autonomous, with no teleoperation involved. Beyond household chores, the Figure 03 is also being prepared for manual jobs such as packaging and parcel delivery.
The hardware revisions prioritize manufacturability and enhanced sensory input. The robot’s hands now feature embedded palm cameras with a wide field of view, ensuring visual awareness even when the main cameras are obstructed, such as when reaching into a cabinet. Additionally, softer and more adaptive fingertips are designed to enable more stable grasps, allowing the robot to handle delicate objects like eggs and playing cards—a notorious challenge for robotic manipulation.
Figure AI has also established a dedicated manufacturing facility, named ‘BotQ,’ to support its ambitious scaling plans. The company plans to ship 100,000 humanoids over the next four years, with the BotQ facility initially capable of producing up to 12,000 units annually. This internal production approach aims to maintain tight control over quality and iteration speed. The Figure 03 also includes safety features like foam to protect against pinch points and fully washable soft textiles covering the body.
The launch follows a successful period of growth for the company, which secured over $1 billion in committed capital last month, raising its post-money valuation to $39 billion. Figure 02 systems are already deployed in a BMW factory, performing 10-hour shifts lifting parts, and the company views Figure 03 as its first truly mass-producible humanoid.
- **AI Core:** Powered by Figure’s proprietary “Helix” vision-language-action (VLA) AI model for real-time reasoning.
- **Hardware Focus:** Redesigned sensory suite, including embedded palm cameras, and softer, adaptive fingertips for delicate manipulation.
- **Application:** Designed for both industrial automation and consumer home use, performing complex chores like folding laundry.
- **Scale:** New ‘BotQ’ manufacturing facility established to meet a goal of shipping 100,000 robots in four years.
Google DeepMind Unveils Advanced ‘Thinking AI’ Models for Robot Reasoning
Google DeepMind has introduced a new pair of artificial intelligence models under its Gemini Robotics project, significantly boosting a robot’s ability to perform complex, multi-step tasks that require high-level reasoning. The new models, Gemini Robotics 1.5 and Gemini Robotics-ER 1.5, are engineered to handle “long-horizon” tasks, bridging the gap toward robots assisting people in real-world environments.
The system operates in a supervisor-worker dynamic. Google Robotics-ER 1.5 acts as the “brain,” a vision-language model (VLM) that processes natural language commands, gathers spatial information, and uses advanced reasoning to formulate a multi-step plan. It then sends these instructions to Google Robotics 1.5, which acts as the “hands and eyes,” a vision-language-action (VLA) model that translates the plan into physical movements.
In a demonstration, the new AI system was shown to be capable of receiving a vague instruction and turning it into a step-by-step plan, using common sense and physics to manage new and unpredictable situations. For example, a robot successfully sorted a selection of fruit onto plates based on shared characteristics and was also able to sort clothes by color, even when the items and bins were moved mid-task, forcing a real-time re-evaluation of the physical space. The same generalist AI model is adaptable to different embodiments, including the Apollo humanoid robot, demonstrating its potential for sophisticated humanoid control.
This development is seen as a major milestone for physical AI, as it allows robots to move beyond simple, one-off commands and execute tasks that require anticipation, planning, and adaptation, which is crucial for the successful deployment of humanoid robots in general-purpose roles.
Yunshen Technology Launches Industry-Grade, All-Weather Humanoid DR02
In a move highlighting the growing specialization within the humanoid robotics sector, China-based Yunshen Technology has announced the release of its first industry-grade, all-weather humanoid robot, the DR02. This robot is specifically engineered to operate in extreme environmental conditions, capable of functioning in temperatures ranging from a frigid -20°C to a scorching 55°C.
The DR02 is positioned to address the rigorous demands of industrial automation, where robots are often required to perform tasks in environments unsuitable or unsafe for human workers. Its all-weather capability suggests applications in construction, outdoor facility inspection, and manufacturing plants with extreme temperature variations. This release aligns with market predictions that the humanoid robotics industry is expected to enter the mass production phase in the coming year.
Analysis: China’s Robotics Sector Leverages Vertical Integration and Policy Support
Recent analysis from The Washington Post suggests that China is establishing a significant lead in the global race for embodied intelligence, particularly in the areas of robot hardware and deployment. While the U.S. maintains an advantage in advanced AI chips and foundational software research, China’s comprehensive supply chain and aggressive deployment strategy are creating a powerful feedback loop for rapid development.
Key factors contributing to this acceleration include vertical integration, where many Chinese startups produce critical components like actuators, sensors, and batteries in-house or procure them domestically. This allows for rapid prototyping and fine-tuning of robots, leading to more affordable and scalable products. For example, Unitree recently released its R1 humanoid for under $6,000, drastically lowering the entry price for advanced robotics.
The Chinese government’s strong policy support and the country’s high-volume factory deployments are spurring the extensive data collection necessary to train the robots’ AI models at scale. In 2024, China installed nearly 300,000 industrial robots, surpassing the rest of the world combined. This large-scale, real-world deployment generates massive amounts of operational data, which is essential for developing the next generation of capable, general-purpose humanoid robots.
“If embodied intelligence — think AI-powered robots that can navigate the real world — is the next frontier of AI, then China appears poised to dominate.” – The Washington Post
The combination of a robust, localized supply chain, competitive pricing, and a national strategy focused on addressing labor shortages through automation is solidifying China’s position as a dominant force in the physical AI landscape.