Launched in July 2025 by LIMX Dynamics, the LIMX Oli is a full‑size general‑purpose humanoid robot engineered as a research and development platform for embodied AI. Standing 165 cm tall and weighing approximately 55 kg, Oli‘s full‑size humanoid form factor matches the scale of human environments, enabling direct deployment in industrial, commercial, and research settings without costly infrastructure modifications [9†L3-L6].

Rather than positioning Oli as a finished product, LIMX Dynamics designed it as an open, modular platform aimed at AI researchers, robotics developers, system integrators, and innovation‑driven enterprises. The robot offers three distinct versions—Lite, EDU, and Super—each built on the same modular hardware‑software architecture but with different feature sets, compute capabilities, and sensor configurations, allowing users to select the version that best fits their project requirements, budget, and development goals [13†L1-L8].

Full‑Size Human‑Scale Design for Real‑World Environments

Unlike compact or tabletop robots, Oli‘s 165 cm height and 55 kg weight place it at a true human scale. This form factor allows the robot to work alongside human personnel across diverse real‑world settings:

• 165 cm / 5′5″ height; 55 cm / 21.7″ shoulder width; 70 cm / 27.6″ arm length

• 31 active degrees of freedom (excluding end‑effectors), distributed for human‑like motion:

  • 7 DoF per arm for complex manipulation tasks

  • 6 DoF per leg for stable bipedal locomotion and terrain adaptation

  • 3 DoF in the waist for coordinated upper‑lower body movements

  • 2 DoF in the neck for flexible head positioning and visual tracking [13†L2-L8]

The Super edition is scaled up to 175 cm / 5′9″ height, 60 cm shoulder width, 80 cm arm length, and 43 total DoF [13†L6-L7].

Deliberate Degree‑of‑Freedom Distribution for Whole‑Body Loco‑Manipulation

Oli‘s DoF distribution is designed for whole‑body loco‑manipulation (coordinated movement and manipulation), a requirement that sets higher‑capability humanoids apart. Each joint provides a wide range of motion:

Dual‑Compute Architecture and Performance

Oli separates motion control and perception tasks onto independent compute units—a design choice that eliminates real‑time interference between the robot‘s “movement” and “sensing” pipelines. Each version offers different compute capabilities:

• Oli Lite & EDU : RK3588 / 8GB RAM / 64GB storage – handles locomotion control tasks

• Oli Super : NVIDIA Jetson Orin NX (157 TOPS) / 16GB RAM / 1TB storage – for vision, LiDAR processing, and AI inference [14†L16-L18]

• End‑effector options: Humanoid hand model, parallel two‑finger gripper (1 DoF), five‑finger dexterous hand (6 DoF) [13†L8-L10]

Quick‑Swap Battery and Optimized Cooling

Oli is equipped with a 9,500mAh quick‑swap drawer‑type battery module (two units per robot), offering approximately 1.5–2 hours of runtime depending on the edition and workload. The Super edition features an enhanced main control area cooling system to maintain high‑performance operation during extended runtime and intensive computation. The 58.8V 10A charger allows full recharging in a short period [13†L21-L23] [14†L11-L15].

Comprehensive Hardware Openness for Customization

Oli features a modular hardware design that supports flexible component swapping:

• End‑Effectors: Interchangeable options include humanoid hand models, parallel two‑finger grippers (1 DoF), and 6‑DoF five‑finger dexterous hands, with self‑developed two‑finger and three‑finger grippers available as optional upgrades [13†L8-L10]

• External Devices: Supports integration of microphones, cameras, tactile sensors, additional IMUs, and LiDAR systems for expanded perception [12†L19-L23]

• Communication Interfaces: WiFi 6, Bluetooth, USB3.0/3.2 (expansion dock supported), Gigabit Ethernet (RJ45, 2 units), Type‑C (expansion dock supported), and 5G [14†L11-L12]

• Power Interfaces: 24V 5A (2 or 4 units depending on edition) and 12V 5A (2 or 4 units) for powering external modules [14†L12-L13]

Flexible Edition Options

LIMX Oli is available in four editions, each designed for specific user needs across R&D, commercial deployment, and industrial applications:

Open Software Ecosystem for Full Development Access

Oli is a platform that researchers can actually control: SDK interfaces are fully open, with access to state data, sensor feeds, low‑level joint control, and high‑level task scheduling. Full Python development is supported, along with a modular SDK that opens both upper‑layer (task planning) and lower‑layer (joint‑level) APIs, enabling use cases from teleoperation system construction to imitation learning data pipelines without modifying the underlying architecture. Clear and complete URDF files ensure smooth simulation‑to‑reality (Sim2Real) transfer. Motion control training code is provided for both X86 and Orin deployment. Oli supports major simulation platforms including NVIDIA Isaac Sim, MuJoCo, and Gazebo. [8†L32-L36] [9†L24-L41]

Smart Mobility and Runtime Management

• Walking Speed: Maximum 5 km/h for stable bipedal locomotion [14†L15-L16]

• Load Capacity: Maximum single‑arm load of 3 kg, sufficient for handling common objects in research and light industrial applications [14†L15-L16]

• Smart Monitoring: Status monitoring and light interaction systems keep operators informed of robot health and operation status [14†L19-L20]

• OTA Updates: Motion libraries and controller modules are delivered via OTA updates, enabling fast iteration and remote deployment [9†L34-L40]

COSA System – The Next Generation Embodied AI OS

In January 2026, LIMX Dynamics launched COSA (Cognitive OS of Agents), a physical‑world native agentic operating system that serves as the “brain” and “nervous system” for Oli. COSA integrates high‑level cognition with whole‑body motion control, enabling the robot to “think, move, and work while thinking.” Built on a three‑layer architecture, COSA provides [16†L2-L25]:

• Bottom Layer: Whole‑body Control Foundation Model for stable locomotion and balance

• Middle Layer: High‑level skills with environmental perception and adaptation (navigation, obstacle avoidance, grasping, stair climbing, etc.)

• Top Layer: Autonomous cognition and decision‑making, including natural language semantic understanding and intent recognition

With COSA, Oli can process complex instructions such as “bring two bottles of water to the front desk” or “find the red package and deliver it to the workshop,” autonomously decomposing long‑horizon tasks and dynamically adjusting task priorities based on environmental changes. It performs simultaneous perception, planning, and manipulation without continuous human guidance. COSA also supports semantic memory and active perception—Oli builds a “worldview” by remembering people, objects, and interaction histories, transitioning from passive input reception to active environmental exploration. The system supports onboard deployment of large language models (LLMs) for natural interaction and multi‑stream sensor fusion for advanced perception and embodied reasoning. [4†L10-L12] [16†L16-L25]

Safety and Transport

• Safety: Handheld remote controller and handheld emergency stop ensure operator control, with the optional emergency stop add‑on for high‑risk research environments [14†L7-L8]

• Transport: Foldable seated storage design with compact packing, enabling single‑person transport and effortless transition between locations [8†L22-L24]

Applications Across Industries

Global Milestones and Recognition

• Industrial Design Recognition: Oli‘s clean, full‑scale form has drawn attention for its professional design—built for workplace functionality rather than speculative “future‑living” showcases

• Large‑Scale Deployment Demonstration: In January 2026, LIMX demonstrated 18 Oli humanoids activating simultaneously, exiting shipping containers, standing up, and moving in coordinated formation without human commands, representing a significant step toward scalable industrial deployment of humanoid fleets [15†L3-L35]

• Rugged Outdoor Validation: Oli demonstrated stable locomotion across sand, rocks, boards, and debris, showing its viability for inspection and patrol tasks beyond climate‑controlled labs [4†L13-L19]

• Autonomous Path Planning: Running the COSA system, Oli can dynamically reprioritize and replan tasks in real time—such as delivering water to a front desk while handling an unexpected parcel collection request—navigating obstacles and manipulating objects without continuous human guidance [10†L36-L42]

• Climbing Validation: Oli successfully climbed 30 curved steps (10 meters vertical ascent) in a high‑wind outdoor environment at the Guangzhou Tower Robot Ascending Challenge, winning the Bipedal Humanoid Group Championship and Best Precision Control Award [20†L2-L9]

• Conference Appearance: Oli opened the 2025 Guangdong‑Hong Kong‑Macao Greater Bay Area AI and Robotics Industry Conference in Guangzhou with a flawless bow, showcasing its refined motion control to industry leaders [21†L4-L6]

• Industrial Validation: LIMX has secured deployment commitments from industrial partners, with early units destined for factory and warehouse trials, including parts sorting, palletizing, and material handling tasks

Available End‑Effectors and Add‑Ons (Optional)

• Humanoid Hand Model: Standard cosmetic hand for human‑like appearance

• Parallel Two‑Finger Gripper (1 DoF): Robust grasping for a wide range of objects in research and logistics

• Five‑Finger Dexterous Hand (6 DoF): Advanced manipulation for complex, human‑like grasping and fine manipulation tasks

• Self‑Developed Two‑Finger Gripper (1 DoF): Proprietary gripper with optimized performance for specific tasks

• Self‑Developed Three‑Finger Gripper (4 DoF): Intermediate‑complexity manipulation for applications where five fingers are excessive but two are insufficient

• External Modules: Seamless integration of voice systems, vision sensors (additional cameras, LiDAR units), tactile sensors, additional IMUs, and other custom modules

• Handheld Controller: Standard remote controller for manual operation

• Handheld Emergency Stop: Operator safety control for high‑risk research and teleoperation environments

• Computing Backpack (Oli Super): Additional compute capacity for extreme AI workloads and advanced perception pipelines

Why Choose LIMX Oli?

• Full‑Size Human‑Scale Platform – 165 cm, 55 kg form factor designed to operate seamlessly in human‑built environments without costly retrofitting

• True Open Architecture – End‑to‑end modular design: swap components, add sensors, and customize both hardware and software without roadblocks

• Full Development Access – Complete SDK, ROS 2 compatibility, Python support, and access to low‑level control for robotics researchers and AI engineers

• Versatile Edition Options – Three editions allow matching capability and budget to project needs, from entry‑level research to advanced AI development

• Scalable Deployment – 18‑unit autonomous coordination demonstration proves readiness for industrial‑scale humanoid fleets

• Proven Outdoor Performance – Validated across sand, rocks, debris, stairs, and outdoor climbing challenges – not just a lab platform

• Growing Global Ecosystem – Dedicated development documentation, OTA support, and expanding industrial partnerships in logistics, warehousing, and manufacturing