Kinetix AI Unveils KAI: A 115-DoF Humanoid Aiming for "Physical Intelligence"

Kinetix AI (also known as Kai Robotics) launched KAI, a humanoid with 115 degrees of freedom
The platform features a 36-DoF dexterous hand and a full-body tactile skin system with 18,000 sensors capable of detecting 0.1N of pressure.
KAI is powered by semi-solid-state battery technology, similar to the energy solutions being pursued by XPENG Robotics.
The company utilizes a unique "first-person" data strategy via the KAI Halo, a head-mounted device used to capture human movement and environment data.

Editor's Note

This article has been updated to correct inaccuracies regarding the team’s background, projected pricing, and the anticipated production timeline.

On April 26, 2026, Shenzhen-based startup Kinetix AI (also operating as Kai Robotics) held its "GIFTED" press conference to unveil KAI, a full-sized humanoid robot designed to bridge the gap between mechanical automation and human-centric service.

Two KAI humanoid robots are seated in modern armchairs on a dimly lit stage during a press conference. The robot on the left is dressed in a white form-fitting suit, while the one on the right wears a black suit. They are positioned as if in conversation, flanking a small wooden side table with a minimalist white sculpture. — Two KAI humanoids appearing on stage at the "GIFTED" press conference in Shenzhen. The launch featured the robots introducing themselves through a simulated dialogue to showcase their communicative and social potential. Image: Kinetix AI

Record-Breaking Kinematics and Dexterity

The most striking technical specification of the KAI platform is its 115 degrees of freedom (DoF) across the entire body. For context, contemporary rivals typically operate with 20 to 45 DoF. This high degree of articulation allows for a range of motion that closely mimics human flexibility, including specific capabilities such as:

Shoulder movement: -20° to 0° lifting and -15° to 0° embracing.
Torso flexibility: Waist bending from -15° to 75°.
Cervical range: Neck rotation from -15° to 50°.

The robot’s hands are equally ambitious, featuring 36 DoF per hand (22 active and 14 passive joints). The passive joints act as mechanical buffers that allows the hand to conform to objects and absorb impacts without immediate computational overhead—a feat that the company says is a critical safety feature for domestic use.

Sensors and Safety: Feeling the World

To ensure safe operation in unconstrained environments, KAI is wrapped in a synthetic tactile skin containing 18,000 "tactels" or sensing points. This system allows the robot to detect forces as light as 0.1N, enabling "haptic-aware manipulation."

Stay ahead in humanoid robotics

One email a week: the launches, funding and research that mattered. Join the researchers, founders and technologists following the field.

Safety is further bolstered by the use of a 1.7kWh semi-solid-state battery. This choice mirrors recent trends in the Chinese robotics sector, including developments at XPENG Robotics, as semi-solid-state chemistry significantly reduces the risk of thermal runaway compared to traditional lithium-ion packs.

A KAI humanoid robot with a sleek silver faceplate and a white textured body suit walks forward while carrying two brown paper shopping bags in its right hand. The background shows a blurred domestic or retail setting with wooden shelving. — KAI demonstrating its 20 kg dual-arm payload capacity by carrying shopping bags. Kinetix AI is positioning the 115-DoF platform as a general-purpose helper for domestic and retail environments. Image: Kinetix AI

The KAI World Model and Data Strategy

Intelligence for the robot is driven by the KAI World Model, a closed-loop system consisting of Base, Action, and Evaluation modules. According to the company, KAI uses this architecture to predict environmental changes and assess the safety and stability of its own "candidate trajectories" before execution.

To solve the industry-wide "data bottleneck," Kinetix AI introduced the KAI Halo, a lightweight, head-mounted terminal. Human operators wear the device during their daily routines, capturing first-person video, poses, and environmental point clouds. They claim that this "first-person experience" provides a more diverse and natural dataset than traditional motion-capture setups.