← Back to comparator
Unitree H2 Plus
Unitree Humanoid Specs-based review

Unitree H2 Plus NVIDIA Isaac GR00T reference

Unitree's NVIDIA-powered research humanoid — 75 DOF, Jetson Thor, Isaac GR00T, due late 2026

N/A*
Not Rated Yet
Price To be Confirmed
Released 2026
Partial score — based on publicly available specifications. Hands-on subcategories pending physical testing.
Visit Manufacturer →

Overview

The Unitree H2 Plus is the humanoid robot at the centre of Unitree's partnership with NVIDIA, announced on 1 June 2026 as an NVIDIA Isaac GR00T reference design for academic robotics research. It pairs a 31-DOF Unitree H2 body with dual Sharpa Wave five-finger tactile hands (22 DOF each) for 75 total degrees of freedom, and runs on NVIDIA's Jetson AGX Thor T5000 — a Blackwell GPU delivering 2,070 FP4 teraflops with 128 GB of unified memory. Standing roughly 1.80 m and weighing about 68 kg, it offers up to 120 N·m of arm torque, 360 N·m of leg torque, a 7 kg rated arm payload (15 kg peak) and around 3 hours of runtime from a 0.972 kWh battery. The reference platform bundles NVIDIA's Isaac GR00T open software and models on top of Unitree's ROS2 ecosystem. The H2 Plus is scheduled for availability in late 2026; pricing has not been announced and RobotTesters has not tested a unit — this is a specs-based preview.

Score Breakdown

Scores not available yet — this robot's price has not been officially confirmed.

Available skills

We haven't catalogued specific skills for the Unitree H2 Plus yet. Our crawler tracks UniStore and public GitHub repos that explicitly target this robot; as the ecosystem grows, new motions, behaviours and SDKs will appear here automatically.

Technical Specifications

Height ~1.80 m (≈6 ft)
Weight ~68 kg (150 lb)
DOF 75 (31 body + 2×22 hands)
Hands Sharpa Wave 5-finger tactile (22 DOF each)
Arm Torque Up to 120 N·m
Leg Torque Up to 360 N·m
Arm Payload 7 kg rated / 15 kg peak
Battery Life ~3 h (15 Ah, 0.972 kWh)
Sensors Head stereo cam 140°×102°, wrist cams, IMU
Compute NVIDIA Jetson AGX Thor T5000 (Blackwell, 2,070 FP4 TFLOPS, 128 GB)
AI NVIDIA Isaac GR00T
Connectivity Wi-Fi 6, Bluetooth 5.2, Ethernet, USB

Pros & Cons

Strengths

  • 75 degrees of freedom — 31 in the body plus dual Sharpa Wave five-finger hands (22 DOF each), among the most dexterous humanoids announced
  • NVIDIA Jetson AGX Thor T5000 onboard: Blackwell GPU, 2,070 FP4 teraflops and 128 GB unified memory — flagship on-robot AI compute
  • Ships as an open NVIDIA Isaac GR00T reference design — foundation models, simulation and SDK out of the box
  • Tactile five-finger hands plus head stereo and wrist cameras geared for close-range manipulation
  • Backed by the Unitree ROS2 ecosystem and a direct NVIDIA–Unitree research partnership

Weaknesses

  • Not on sale until late 2026 — no price announced, so value for money cannot be assessed
  • Aimed at academic and robotics research, not consumer buyers
  • Max speed and full battery runtime under load not disclosed for the H2 Plus configuration
  • All capability figures come from the manufacturer announcement — not independently tested
Visit Manufacturer →
Specs-based analysis — score derived from manufacturer data. How we score

Compare with others