Skip to content
  1. Home
  2. News
  3. Unitree releases the UnifoLM-WLA-1.0-Base model weights and fine-tuning code
News · Other

Unitree releases the UnifoLM-WLA-1.0-Base model weights and fine-tuning code

Unitree released the base weights and fine-tuning code of UnifoLM-WLA-1.0, a 6-billion-parameter model trained on about 2,500 hours of real-robot data, on 28 Sep 2026 under the Apache 2.0 license.

Date
primary source published
Type
Other
Primary source
github.comDataset
20 Oct 2026Unitree Robotics (code repository) · dated 28 Sep 2026
Added
Checked

Unitree’s UnifoLM-WLA repository records the release on 28 Sep 2026: “we released the UnifoLM-WLA-1.0-Base and fine-tuning code.” The base weights are on Hugging Face, and the code is under the Apache License 2.0.

The repository describes a model with 6 billion parameters trained on about 2,500 hours of real-robot data, covering 64 tasks across tabletop and whole-body manipulation. Earlier steps of the same release are dated 11 Sep 2026 (weights of UnifoLM-ER-1 and UnifoLM-ER-Flow) and 20 Sep 2026 (model modules and action expert training code). The project page lists its real-robot tests under the heading “Unitree G1.”

The results on the project page are Unitree’s own. The release adds an open model for the G1 platform; it does not change the price or status on the Robopear G1 card.

Sources

Sources of this item, primary first
RoleDocumentLink / checked
Primaryunitreerobotics/unifolm-wlaUnitree Robotics (code repository) · dated 28 Sep 2026
github.comDataset
20 Oct 2026
SecondaryUnitree Robotics (model page)dated 28 Sep 2026
20 Oct 2026
SecondaryUnifoLM-WLA project pagedated 28 Sep 2026
20 Oct 2026

Related cards

ModelUnitree G1

Model card · Unitree

CompanyUnitree

Company card · China

This item does not change the cards above. A price or status on a card moves only after the data check confirms it with a source and a date: verification policy. All news · RSS feed.