VLA GalleryAll modelsGallery

VLA Gallery / All models / Industry models

UnifoLM-WMA-0

Unitree · 2025-09

기타Industry modelsNot disclosedvideo world model

Unitree's open framework combining a world model and action generation, demonstrated with arm tabletop manipulation.

View architecture diagram in the galleryGitHub ↗

Paper

No paper

Overview

Category
Industry models
Type
기타
Released
2025-09
Organization
Unitree
Open weights
Yes
Parameters
Not disclosed

Architecture

Backbone LLM
Not disclosed
Vision encoder
—
Action head
world-model-action (details not disclosed) (video world model)
Chunk
Not disclosed
Control rate
Not disclosed
Architecture (original summary)
world-model-action framework (decision-making·simulation modes); detailed architecture not disclosed

Training

Pretraining accelerators
Not disclosed
Pretraining duration
Not disclosed
Pretraining steps & batch
Not disclosed
GPU/TPU-hours
Not disclosed
Fine-tuning
Not disclosed
Data type
Public robot data (Open-X, public Unitree datasets)
Data scale
Not disclosed
Main datasets
Open X-Embodiment, 5 public Unitree datasets
Tasks & evaluation
Unitree arm tabletop manipulation demonstrations; quantitative benchmarks not disclosed
Robot
Unitree Z1 arm (single·bimanual), G1 (Dex1 gripper)

Inference

Inference device & speed
Not disclosed
Inference memory
Not disclosed
Weights bf16 (calc.)
Not disclosed
Where it runs
Not disclosed

Reinforcement learning & base

Reinforcement learning
Not disclosed · Training method not disclosed
Base control
Fixed · No base·locomotion control (Z1 arm·G1 Dex1 tabletop manipulation)

Notes

Key features
Unitree's open framework combining a world model and action generation, demonstrated with arm tabletop manipulation. Base was fine-tuned on Open-X, and Dual on 5 public Unitree datasets.
Notes
Information limited since only README was checked (paper unverified). Weights·code released 2025-09-15. Type set to Other since VA/VLA status is unverified.
Verification
Confirmed

Other models in this category