Robbyant VLA pretrained on 20,000 hours from nine dual-arm configurations.
Access and availability
- Code
- Public access
- Model files
- Public access
Code license
- LingBot-VLA·Apache-2.0
SourceVerified on · 2026-10-27
Training / fine-tuning code
- Access
- Public access
Verified on · 2026-10-27
- Code license
- Apache-2.0
Model files
| File | Access |
|---|---|
Getting started
- Deployment requirements
- Official environment: Python 3.12.3 and PyTorch 2.8. The depth variant additionally uses MoGe-2; follow the matching checkpoint configuration.
At a glance
- Model type
- Vision-language-action
- Backbone
- Qwen2.5-VL-3B; optional MoGe-2 depth features
- Released
- January 27, 2026
Technical details
Functional purpose
Action & control
Weights publication date
2026-01-27
Training data
LingBot-VLA pretraining mixture
Pretraining20,000 hours of real-robot data from nine dual-arm configurations. This describes the reported pretraining corpus, not a downloadable dataset.
Before you use this version: Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.
Submit a source or correction