Robot Assessments
Structured ZeroGantry assessments of deployed and near-production robots, with pros, cons and an editorial verdict. Each is based on published specifications and deployment reports, not hands-on testing. Assessments link into Robot Reference and Compare for side-by-side specs.
35 assessments published. Reviews RSS · Robot Reference · Methodology
- Unitree H2 assessment — Unitree Robotics — The H2 delivers respectable full-size specs at a disruptive price point but arrives half-finished for serious engineering work. Its 31-DoF architectur
- Booster K1 assessment — Booster Robotics — The Booster K1 delivers credible agility for its size class, routinely recovering from disturbances and posting competitive RoboCup results with its 2
- Apptronik Apollo 2 assessment — Apptronik — Apptronik Apollo 2 prioritizes data generation over polished autonomy, positioning it as a capable testbed rather than a ready production worker. Its
- Spot Quadruped assessment — Boston Dynamics — Spot remains the reference commercial quadruped for industrial inspection, its mature 12-DoF platform and Orbit-managed Autowalk delivering repeatable
- Atlas (Electric Humanoid) assessment — Boston Dynamics — The electric Atlas sets the benchmark for dynamic humanoid mobility with its 56-DoF electric architecture and self-maintaining battery system, deliver
- HUBO-2 Plus assessment — Rainbow Robotics — The HUBO-2 Plus remains a capable academic research platform rooted in KAIST's proven bipedal lineage, yet its 2008-era architecture shows its age aga
- XPeng IRON (Gen 2) assessment — XPeng — XPeng's IRON Gen 2 represents a credible automotive-to-robotics transfer play, leveraging three Turing chips and VLA stack for onboard autonomy that f
- Reachy assessment — Pollen Robotics — Reachy delivers a genuinely hackable research torso that prioritizes accessibility over raw capability, with its Python-first SDK and ROS2 stack accel
- Astribot S1 assessment — Astribot (Stardust Intelligence) — The Astribot S1 stands out for its cable-driven arms that deliver superhuman speed and sub-millimeter precision, validated in academic papers and vira
- 1X NEO assessment — 1X Technologies — 1X NEO stands out for its consumer-oriented safety focus and lightweight construction, making it one of the few humanoids plausibly suited for unstruc
- Sony aibo (ERS-1000) assessment — Sony — The ERS-1000 remains Sony's most refined quadruped companion, with its 22-DoF architecture and cloud-driven learning delivering genuine emotional enga
- NAO assessment — Aldebaran / SoftBank Robotics — NAO remains the most accessible entry point for humanoid research and STEM education, but its 2018-vintage hardware now feels dated against newer plat
- Pepper assessment — SoftBank Robotics — Pepper's 20-DoF expressive upper body and emotion-aware perception delivered measurable engagement lifts in early retail pilots, yet its 3 km/h wheele
- PARO assessment — AIST / PARO Robots — PARO remains the clinical benchmark for non-pharmacological dementia intervention, its sensor-driven responsiveness and FDA pedigree delivering measur
- AGIBOT Expedition A3 assessment — Agibot — The Expedition A3 excels as a lightweight performance platform whose 0.218 kW/kg density and dual-battery architecture deliver repeatable dynamic moti
- TIAGo Pro assessment — PAL Robotics — The TIAGo Pro stands as a mature research workhorse whose series-elastic dual arms and ROS 2 architecture deliver reliable data for HRI and embodied A
- Galbot G1 assessment — Galbot — The Galbot G1 earns its place as the first robot to secure a licensed pharmacy operating permit in China and run continuous shifts at FamilyMart, prov
- Hyodol assessment — Hyodol Co. — Hyodol delivers measurable scale in South Korean eldercare with over 12,000 units in solitary homes, leveraging ChatGPT for contextual conversation an
- Toyota T-HR3 assessment — Toyota — Toyota's T-HR3 remains a 2017-era research platform whose teleoperation strengths are undercut by minuscule payload and complete absence of autonomy.
- Agibot Expedition A1 assessment — Agibot — The Expedition A1 established Agibot as a credible Chinese humanoid entrant but remains a proof-of-concept platform rather than a production workhorse
- LimX Oli assessment — LimX Dynamics — LimX Oli stands out for its production-ready swarm capability and COSA 0.5 agentic OS, which delivered China's first reported fully autonomous long-ho
- temi assessment — temi Inc. — temi remains a competent telepresence and companion platform whose 5 cm localization and 1 m/s cruise speed suffice for uncluttered residential and sm
- Figure 02 assessment — Figure AI — Figure 02 delivers credible industrial traction with its on-device speech-to-speech stack and proven 16-DoF hands, yet the 5-hour endurance and $150k
- Agibot A2 Ultra assessment — Agibot — A2 Ultra is chasing the “humanoid greeter” job Pepper popularized, with modern walking endurance marketing. The Guinness walk is a clever trust signal
- Walker S2 assessment — UBTECH Robotics — Walker S2 is China’s industrial humanoid bet: high DoF, hot-swap energy, and big domestic order headlines. Treat production claims as directionally im
- Apptronik Apollo assessment — Apptronik — Apollo is the humanoid pitched to people who actually move boxes for a living: payload first, compliance second, viral parkour last. That positioning
- LBR iiwa assessment — KUKA Robotics — LBR iiwa invented the modern sensitive-arm category. Torque-per-axis sensing and 7-DoF kinematics still matter when you are mating fragile connectors
- UR10e assessment — Universal Robots — UR10e remains the default answer when an engineer says “we need a cobot next to a CNC.” It is not the flashiest arm of 2026, and that is the point: do
- ANYmal assessment — ANYbotics — ANYmal is the adult in the quadruped room when the environment can explode. Spot owns general industrial inspection mindshare; ANYmal X owns the hazar
- Tesla Optimus Gen 2 assessment — Tesla, Inc. — Optimus Gen 2 demonstrates credible hardware progress in a 57 kg package capable of real factory tasks, yet its value remains tethered to Tesla's inte
- Unitree Go2 assessment — Unitree Robotics — Go2 is the democratization play Spot never made. The Air SKU at roughly sixteen hundred dollars put a LiDAR-equipped quadruped in dorm rooms and high-
- Unitree G1 assessment — Unitree Robotics — Unitree G1 is the humanoid that actually showed up in university labs. At $13.5–16k it broke the “enterprise only” price wall and created a generation
- Digit assessment — Agility Robotics — Digit has delivered the first verifiable commercial ROI for humanoids, moving 100k+ totes and logging 65k hours where every other platform remains in
- Unitree H1 assessment — Unitree Robotics — Unitree H1 is the value champion of the humanoid market. At under $90k, it's the only humanoid you can actually buy today without a multi-year enterpr
- Fourier GR2 assessment — Fourier Intelligence — Fourier GR2 is the dark horse of the humanoid market — less hyped than Figure or Optimus, but with the best force-control architecture of any robot in