Shanghai Jiao Tong University · robot AI model
WLA-0
China · 2026
Sources retrieved: 19 November 2026 (A, B, C)
- Model type
- Vision-language-action (VLA) World action modelSource and note: Model typeA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026Note. Developer wording: ""We propose world-language-action (WLA) models as a new class of embodied foundation models.""
- Inputs
- Camera Language Robot stateSource and note: InputsA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026
- Access
- Open weights (download)Source and note: AccessB github.com/SJTU-DENG-Lab/WLA · Retrieved 19 November 2026

Specifications
5 of 14 stated 9 not stated
5 fields hidden 4 not-stated fields with a note stay ·
How to read the figuresShow guideClose
- not statedNo source
- 0Reported as zero
- ASource
- A*Source with a caveat (select it)
Select a source marker for source details and notes.
Release1/1 disclosed
Released4 June 2026
Source and note: ReleasedA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026
Input and output2/4 disclosed
Model typeVision-language-action (VLA) World action model
Source and note: Model typeA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026Note. Developer wording: ""We propose world-language-action (WLA) models as a new class of embodied foundation models.""
InputsCamera Language Robot state
Source and note: InputsA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026
What it controlsNot stated
Degrees of freedom controlledNot stated
Size and speed0/2 disclosed
ParametersNot stated
Source and note: ParametersA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026Note. Paper abstract: “Our WLA-0 prototype, with 2B active parameters, achieves 40 ms per inference on an NVIDIA RTX 5090.”
Control rateNot stated
Source and note: Control rateA arxiv.org/abs/2606.05979 · Retrieved 19 November 2026Caveat. Paper abstract reports inference latency rather than control rate: “Our WLA-0 prototype, with 2B active parameters, achieves 40 ms per inference on an NVIDIA RTX 5090.”
Where it runs and what it is built on0/2 disclosed
On robot or cloudNot stated
Base modelNot stated
Openness1/3 disclosed
AccessOpen weights (download)
Source and note: AccessB github.com/SJTU-DENG-Lab/WLA · Retrieved 19 November 2026
Licence documentNot stated
Source and note: Licence documentC raw.githubusercontent.com/SJTU-DENG-Lab/WLA/main/LICENSE · Retrieved 19 November 2026Caveat. The GitHub code repository carries an Apache License Version 2.0, while the HuggingFace model checkpoints carry no model card or weight licence.
Licence typeNot stated
The developer’s own claims1/2 disclosed
Learns new tasksNeeds retraining
Source and note: Learns new tasksB github.com/SJTU-DENG-Lab/WLA · Retrieved 19 November 2026Note. Code repository: ""Jointly trained on 45 seen tasks and cross-embodiment videos of 5 unseen tasks"".
Works across robotsNot stated
Source and note: Works across robotsYes 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.
Notes
SJTU DENG Lab introduced WLA-0 as a world-language-action prototype with code and HuggingFace checkpoints.
Sources
One line per source, not per figure: a source has one retrieval date.
- Aarxiv.org/abs/2606.05979Retrieved 19 November 2026
- Bgithub.com/SJTU-DENG-Lab/WLARetrieved 19 November 2026
- Craw.githubusercontent.com/SJTU-DENG-Lab/WLA/main/LICENSEMain source. Retrieved 19 November 2026



