Comparison controls
Swap robots without losing the table. Green cells mark numeric leaders; text rows show the tradeoffs that need judgment.
Green cells flag numeric leaders, while AI, sensors, and capability rows keep the real buying tradeoffs visible.
Price
Weight
Battery Life
Max Speed
AI
Sensors
Connectivity
Status
Category
Capabilities
| Spec | ||||
|---|---|---|---|---|
| Price | N/A | N/A | $13,500 USD | ¥198,000 CNY |
| Height | 160cm | 165cm | 132cm (standing) | 169cm |
| Weight | 54kg | ≤55kg / 121.3lbs (battery included) | 35kg (with battery) | 69kg |
| Battery Life | About 4 hours per charge (LimX laboratory data; actual runtime varies) | About 2h (lab power-test room; actual data may vary) | ~2 hours | 2 hours (700 Wh swappable battery) |
| Charging Time | About 1 hour | Not disclosed | Not disclosed | Charging supported via standby station; exact charging time not officially disclosedNot safely comparable as a typed unit |
| Max Speed | 5 km/h | 5km/h | >2 m/s trot speed | Not officially disclosed |
| Degrees of Freedom | N/A | N/A | N/A | N/A |
| AI | COSA (Cognitive OS of Agents) + VideoGenMotion (VGM) video-to-motion framework | COSA (Cognitive OS of Agents) — physical-world-native agentic OS | 8-core high-performance CPU (optional NVIDIA Jetson Orin for EDU) | LLM/RAG full-duplex interaction stack with edge deployment, facial recognition, lip-reading, ActionGPT motion generation, HIMUS 3D-SLAM, VectorFlux planning/control, and RTMOF motion control. |
| Sensors | In-house 6-axis IMU, RGB Cameras | Self-developed 6-axis IMU, Head-mounted depth camera, Chest-mounted depth camera, Wrist depth camera mount, Hip depth camera, LiDAR | Depth Camera, 3D LiDAR, 4 Microphone Array, Dual Joint Encoders | LiDAR, RGB-D Cameras, Fisheye Cameras, Six high-definition panoramic cameras, Microphone, Speaker, Force/Torque Sensors, Interactive Screen |
| Connectivity | Wi-Fi 6, Bluetooth | WiFi 6, Bluetooth, USB3.0/3.2 (expansion dock supported), Gigabit Ethernet (RJ45), Type-C (expansion dock supported), 5G | Wi-Fi 6, Bluetooth 5.2 | Remote control, Smartphone control, Computer control |
| Voice Assistants | None | Voice/Text Prompt Interaction | None | None |
| Status | Prototype | Available | Available | Available |
| Category | Humanoid | Humanoid | Humanoid | Humanoid |
| Manufacturer | LimX Dynamics | LimX Dynamics | Unitree | AGIBOT |
Capability matrix
Use this grid when the shortlist is already close and feature gaps become the real tiebreaker.
| Capability | Luna | Oli | G1 | A2 |
|---|---|---|---|---|
| 27 Active Degrees of Freedom | ✓ | — | — | — |
| 3 kg Max Arm Payload | ✓ | — | — | — |
| 31 Degrees of Freedom | — | ✓ | — | — |
| 360° Semantic Perception | — | — | — | ✓ |
| 3D SLAM Navigation | — | — | — | ✓ |
| 5 km/h Max Walking Speed | ✓ | — | — | — |
| Autonomous Cognition | — | ✓ | — | — |
| Bipedal Walking | — | — | ✓ | — |
| Bipedal Walking with Human-like Gait | ✓ | — | — | — |
| Business Consultation | — | — | — | ✓ |
| Calligraphy and Dance Demonstrations | — | — | — | ✓ |
| Choreography and Manual Teaching via Content Development Kit | ✓ | — | — | — |
| Dexterous Hand Interaction | — | — | — | ✓ |
| Dexterous Hands (optional Dex3-1) | — | — | ✓ | — |
| Enterprise Knowledge-Base Q&A | — | — | — | ✓ |
| Exhibition and Guided Presentations | — | — | — | ✓ |
| Facial Recognition and Lip-Reading | — | — | — | ✓ |
| Foldable Design (69cm folded) | — | — | ✓ | — |
| Front-Desk Reception | — | — | — | ✓ |
| Full-Duplex Voice Dialogue | — | — | — | ✓ |
| Hot-swap Supported Drawer-style Battery | ✓ | — | — | — |
| Humanoid Human-Robot Interaction | — | — | — | ✓ |
| Intelligent Obstacle Avoidance | — | — | — | ✓ |
| L4-Level Autonomous Mobility (manufacturer claim) | — | — | — | ✓ |
| Loco-manipulation (walk + manipulate simultaneously) | — | ✓ | — | — |
| Marketing and Customer Service | — | — | — | ✓ |
| Natural Motion Generation from Voice Intent | — | — | — | ✓ |
| OTA Motion Library Updates | — | ✓ | — | — |
| OTA Upgrades | — | — | ✓ | — |
| Object Manipulation | — | — | ✓ | — |
| Quick Battery Replacement | — | — | — | ✓ |
| Research Platform | — | — | ✓ | — |
| Research Platform (Python SDK) | — | ✓ | — | — |
| Rough Terrain Navigation | — | ✓ | — | — |
| Secondary Development (EDU) | — | — | ✓ | — |
| Self-Recovery (get up from lying down) | — | ✓ | — | — |
| Sound-Source Localization | — | — | — | ✓ |
| Swarm Control Support via Content Development Kit | ✓ | — | — | — |
| Video-based Movement Learning (VGM) | ✓ | — | — | — |
Use this flow to get to a clean shortlist fast. The route works best as a dense research workbench, not a marketing page.
Start with robots that solve the same job. Use category anchors like Humanoid, Cleaning, Companions, Research before you worry about micro-spec wins.
Check price, status, battery, sensors, and capabilities before getting lost in minor spec rows.
Once you have a shortlist, hide shared rows. That turns the table from a spec dump into a real decision tool.
A dedicated cleaner versus a home assistant with cleaning features can reveal whether you really need a specialist or just broader household coverage. Compare around the same outcome, not the same marketing language. And when blank-page friction is the problem, use one of the prebuilt quick-start pairings in the workbench above — open one, swap robots, and keep moving.
Turn comparison rows into an actual recommendation — not just more reading.
Not all spec deltas matter equally. Rank your buying criteria before crowning a winner.
The sticker price is only one row. Subscriptions, consumables, and service access often decide the real winner.
Subscriptions
Cloud AI and premium app tiers can make a cheaper robot more expensive over three years.
Consumables
Brushes, pads, filters, and bags flip the value story on high-frequency robots.
Energy & battery
Large batteries and frequent charging reveal everyday operating demands.
Repair & warranty
Local servicing and spare-parts access are the quietest but most important long-term costs.
Specs describe ideal conditions. Your rooms, pets, and Wi-Fi determine what actually performs.
Of 410 robots tracked, 252 are available now. For household tasks, getting a capable robot into daily use today usually beats waiting for incremental improvements. For research or early-adopter use, waiting can make sense if the next model changes the platform story.
If two robots are still neck-and-neck, use softer signals that affect ownership quality more than spec sheets admit.
Comparison FAQ
Short answers to the questions that usually show up once the shortlist is live and the tradeoffs feel real.
