Boston Dynamics Atlas vs PAL Robotics TALOS vs ROBOTIS AI Worker FFW-SG2
Price, availability in your region, size to scale and every spec side by side. Better values are marked; estimates never win.
Quick verdict
Atlas wins 5 of the 8 specs where one robot is clearly better.
5
3
Choose Atlas for
- Degrees of freedom56
- HandsDexterous hands
- Carrying payload30 kg
- Payload per arm20 kg
Choose TALOS for
No outright win on sourced specs.
Choose AI Worker FFW-SG2 for
- Walking speed1.5 m/s
- Runtime6 h
- Battery2,040 Wh
Size to scale
- Adult175 cm
Atlas190 cm
TALOS175 cm- AI Worker FFW-SG2162 cm
All specs
| Spec | AtlasTBA | TALOSOn request | AI Worker FFW-SG2On request |
|---|---|---|---|
| Where you can get it | |||
| United States | Business only | Business only | Ships to you |
| European Union | Not available | Business only | Not available |
| United Kingdom | Not available | Business only | Not available |
| China | Not available | Not available | Not available |
| Japan and Korea | Not available | Not available | Not available |
| The rest of the world | Not available | Business only | Ships to you |
| Overview | |||
| Status | Pilot deployments | Shipping | Shipping |
| Announced | Jan 2026 | Feb 2017 | —Not published |
| Locomotion | Two legs | Two legs | Wheeled base |
| Form factor | Full-size humanoid | Full-size humanoid | Upper body |
| Autonomy | Partly autonomous | —Not published | Partly autonomous |
| Price | |||
| Price type | Not announced | Quote only | Quote only |
| Price note | Not sold publicly; select enterprise early adopters | Request a quote from PAL Robotics; no list price published | Quote only; tariffs included for US quotes |
| Size & body | |||
| Height | 190 cm | 175 cm | 162 cm |
| Weight | 90 kg | 95 kg | 90 kg |
| Vertical reach | 230 cm | —Not published | —Not published |
| Outer shell | —Not published | —Not published | Aluminum, plastic |
| Movement Atlas 1TALOS 0AI Worker FFW-SG2 1 | |||
| Degrees of freedom | 56better | 32 | 25 |
| DoF without hands | —Not published | 30calc. | 23calc. |
| Walking speed | —Not published | 0.8 m/s | 1.5 m/sbetter |
| Actuators | Electric rotary | —Not published | Electric rotary |
| Hands & handling Atlas 4TALOS 0AI Worker FFW-SG2 0 | |||
| Hands | Dexterous handsbetter | Simple gripper | Simple gripper |
| DoF per hand | —Not published | 1 | 1 |
| Touch sensing | Fingers and palms | —Not published | —Not published |
| Carrying payload | 30 kgbetter | —Not published | 6 kg |
| Payload per arm | 20 kgbetter | 6 kg | 3 kg |
| Peak lift | 50 kgbetter | —Not published | 10 kg |
| Battery & power Atlas 0TALOS 0AI Worker FFW-SG2 2 | |||
| Runtime | 4 h | 1.5 h | 6 hbetter |
| Battery | —Not published | 1,080 Wh | 2,040 Whbetter |
| Battery swap | Swaps itself | —Not published | —Not published |
| Charging | Self-docking | —Not published | —Not published |
| Battery swap time | 3 min | —Not published | —Not published |
| Charge time | 90 min | —Not published | —Not published |
| Charges itself | Yes | —Not published | —Not published |
| Sensors | |||
| Cameras | 360° camera view | RGB camera, 1280 x 720 at 30 fps, 60° x 49.5° field of view | Stereolabs ZED Mini head camera, 2 × Intel RealSense D405 hand cameras |
| Depth sensing | —Not published | Depth camera, 640 x 480 at 30 fps, 0.4–8 m range | RGBD cameras, 2 × LakiBeam 1 LiDAR (270°, up to 25 m) |
| Audio | —Not published | 5 W speaker, English text to speech | —Not published |
| Compute & AI | |||
| Onboard computer | —Not published | 2 × Intel Core i7 (COM Express Type-6), Ubuntu LTS with RT Preempt | NVIDIA Jetson AGX Orin 32GB |
| AI model | Large behavior models; Gemini Robotics models in joint research with Google DeepMind | —Not published | —Not published |
| Remote operation | Yes | —Not published | —Not published |
| Connectivity | —Not published | Wi-Fi 802.11 a/b/g/n, Ethernet, EtherCAT | Ethernet, Wi-Fi router |
| Software | |||
| SDK | —Not published | —Not published | Open SDK |
| Languages & frameworks | —Not published | —Not published | ROS 2, Python, C++ |
| Simulation models | —Not published | Gazebo and URDF model (open source) | Gazebo, MuJoCo, Isaac Sim/Lab |
| Safety & compliance | |||
| Dust & water rating | IP67 | —Not published | —Not published |
| Intended use | Industrial material handling: part sequencing, machine tending, order fulfillment | Humanoid robotics research | Industrial manipulation, imitation learning and physical AI research |
| Support | |||
| Service | Field-replaceable limbs in under 5 minutes; customer self-repair certification | —Not published | —Not published |
A check marks the better value. Height and weight are context, not wins. Values marked "est.", "reported" or "calc." never count as wins. How we compare




