humanoid.guide introduces Unitree Robotics’ three-finger robot hand, Dex3-1

humanoid.guide introduced Unitree Robotics’ three-finger robot hand, Dex3-1, on November 4, 2026. According to the outlet, it has a thumb and two fingers and is designed primarily for the company’s G1 humanoid.

According to humanoid.guide, Dex3-1 is designed primarily for the G1 and weighs 0.71 kilograms. Selecting an end effector for a robot requires checking its connection to the main body and its measured performance on the intended tasks. The ratings published by the outlet do not indicate lifting performance, so it is important to distinguish what the numbers represent.

Dex3-1’s distinguishing feature is that it has a thumb and two fingers rather than five fingers. Its design, with a more limited finger arrangement, highlights the need to consider not only the number of fingers but also contact points with objects and how movements are performed when assessing an end effector’s configuration. However, the introduction does not provide experimental conditions using objects or task results, so it is not possible to determine which tasks this configuration is suited for. Even when considered alongside an article introducing the Dex2/5, which we also published that day, the information available only confirms that multiple articles have introduced Unitree Robotics robot hands. No information is provided on the mechanical design, uses or generational relationship between the Dex3-1 and Dex2/5, and their product-line differences should not be inferred from their names or the proximity of their publication dates. The claim that the hand is intended for the G1 highlights the need to consider its specifications in combination with the robot body. Its dimensions and weight offer an indication of its size as a component, but the article does not specify the mounting method, connection to the body or motion-control specifications. The description that most joints are direct-drive and one is gear-driven is also insufficient to assess speed, force, durability or control precision. The Hand Score of 3/10, Degrees of Freedom score of 2/5 and Strength score of 1/5 are assessments by humanoid.guide, not measured values. The outlet explicitly states that the ratings do not represent the number of joints or lifting performance, so they cannot be taken as direct indicators of the product’s practicality or task capabilities. Assessing actual performance requires checking the rating methodology and criteria, as well as the objects used, required contact force and conditions of grip tests. Nor can Dex3-1’s hardware be linked to a specific implementation of skills development based on an article on Unitree’s development of “Rehearsal” to support skill improvement in humanoids. The materials provided do not say how Dex3-1 would be used in control or learning procedures. The next points to verify are its specific connection and control specifications with the G1, gripping conditions in actual tasks, and the product’s price and availability. Without this information, it is too early to judge its suitability for deployment based on the introduction’s dimensions and ratings. See the original source for details.