Robot AI model
Wall-OSS-0.5
Wall-OSS-0.5 co-trains action and multimodal prediction using Qwen2.5-VL-3B-Instruct. The report labels the model 4B but describes the full architecture as over 4B, so an exact total is not assigned; independently stated checkpoint license terms were not established.
Records last verified Oct 13, 2026 · Send a correction
- Organization
- X Square Robot
- Type
- VLA
- Released / introduced
- May 29, 2026
- Public weights
- Public weights
- Licence / scope
- Not established
- Parameters
- Not disclosed
- Base model
- Qwen2.5-VL-3B-Instruct
- Humanoid association
- None reported
- Inputs
- vision, language, proprioception
- Outputs
- continuous robot action chunks
Release dates can mark an announcement or paper introduction. Public checkpoints, code licences and weight terms are distinct; follow the publisher’s links for the stated scope.
Disclosed training data
- RoboMIND v1 →
- RoboMIND v2.0 →
- AgiBotWorld Beta →
- Galaxea Open-World →
- DROID →
- BridgeData V2 →
- Self-collected data, other open robot trajectories and grounded multimodal data
More robot AI models
Evidence
Sources
- [s1] Wall-OSS-0.5 Technical Report ↗
X Square Robot · paper · published May 29, 2026 · accessed Oct 13, 2026
- [s2] Wall-OSS-0.5 official model card ↗
X Square Robot · official · accessed Oct 13, 2026
- [s3] Wall-OSS-0.5 report: parameter scope, initialization and training mixture ↗
X Square Robot · paper · published Jun 1, 2026 · accessed Oct 13, 2026