Skip to content
embodied.guide

X-HumanoidChina2026

Pelican-Unify

Visit website

Key figures as stated by the developer

3 of 7 key figures stated
  • Model type ⁠A world action model
  • Inputs ⁠A camera language robot state
  • Base model ⁠B Qwen3-VL-4B-Instruct + Wan2.2

Not stated: Controls · Access · Parameters · Licence

Full specifications

4 of 14 fields stated
The letter points to the source at the foot of the page. Fields without a figure are dashed: the developer does not state them — that is not the same as zero, and not the same as no.
Field Value Source and caveats
Release
Released14 May 2026

[A] "Submitted on 14 May 2026" ⁠C
Input and output
Model typeworld action model

X-Humanoid's model page is headed "UNIFIED WORLD-ACTION MODEL". ⁠A

Inputscamera language robot state

X-Humanoid says the model combines multi-view images, language instructions, visual history and robot actions (translated from the Chinese). The paper: it "bundles the action history, language instruction, environment observations, and robot proprioceptive state". ⁠A

Controlsnot stated

Note The paper names the output as "the executable action chunk", not a robot body part. ⁠B

Degrees of freedom controllednot stated
Size and speed
Parametersnot stated
Control ratenot stated
Where it runs and what it is built on
On robot or cloudnot stated
Base modelQwen3-VL-4B-Instruct + Wan2.2

Paper: "A VLM backbone initialized from Qwen3-VL"; "Unified Future Generator initialized from Wan2.2". Also: "Qwen3-VL-4B-Instruct (the 4B base model we build upon)". ⁠B
Openness
Accessnot stated
Licence termsnot stated
Licencenot stated
The developer’s own claims
Learns new tasksnot stated
Works across robotsnot stated

Yes only if the developer itself says the model runs on different robot models, for example robots from two manufacturers or two of one manufacturer’s models. Several units of the same robot do not count.
How to read the figures
50HzA
The letter in the grey square after a figure points to the source the figure came from. Pages with several sources carry several letters.

What the data labels mean

Notes

  • X-Humanoid calls Pelican-Unify a unified foundation model for embodied intelligence that closes the loop of thinking, imagining and acting (translated from the Chinese).
  • The developer paper builds the language model part on Qwen3-VL-4B-Instruct and the future generator on Wan2.2.
  • The developer paper tests the model on a UR5e robot arm and a Tienkung humanoid robot.
Versions and notes in the key figures (4)

Model type

X-Humanoid's model page is headed "UNIFIED WORLD-ACTION MODEL".

Controls

Note The paper names the output as "the executable action chunk", not a robot body part.

Inputs

X-Humanoid says the model combines multi-view images, language instructions, visual history and robot actions (translated from the Chinese). The paper: it "bundles the action history, language instruction, environment observations, and robot proprioceptive state".

Base model

Paper: "A VLM backbone initialized from Qwen3-VL"; "Unified Future Generator initialized from Wan2.2". Also: "Qwen3-VL-4B-Instruct (the 4B base model we build upon)".

Sources

One line per source, not per figure: a source has one retrieval date. The letter beside a figure points to the line it came from.