RobotEra · robot AI model
VPP2
China · 2026
Sources retrieved: 2 November 2026 (A, B, C, D, E, F)
- Model type
- World action modelSource and note: Model typeD github.com/roboterax/video-prediction-policy-2 · Retrieved 2 November 2026Note. RobotEra's own term in the code repository: “a world-action model (WAM)”.
- What it controls
- Arms and handsSource and note: What it controlsA github.com/roboterax/video-prediction-policy-2/blob/main/docs/robodojo.md · Retrieved 2 November 2026Note. RoboDojo guide, action format of the dual-arm setup: “State/action is dual-arm absolute EE16”.
- Inputs
- Camera Language Robot stateSource and note: InputsA github.com/roboterax/video-prediction-policy-2/blob/main/docs/robodojo.md · Retrieved 2 November 2026Note. RoboDojo guide: “The policy reads RGB, robot state and instructions through the standard interface.”
- Access
- Open weights (download)Source and note: AccessC huggingface.co/Haodong082399/VPP2 · Retrieved 2 November 2026Note. Hugging Face model page, listed as not gated: “Official VPP2 weights for zero-shot video prediction, RoboDojo and LIBERO.”
- Base model
- Wan2.1-I2V-14BSource and note: Base modelB arxiv.org/html/2610.10270 · Retrieved 2 November 2026Note. Paper: “VPP2 builds upon the pretrained Wan2.1-I2V-14B model”.
Specifications
7/14 specifications disclosed 7 not stated
1 field hidden 6 not-stated fields with a note stay ·
How to read the figuresShow guideClose
Not statedThe developer publishes nothing on it
Select a source marker for source details and notes.
Release1/1 disclosed
Released7 October 2026
Source and note: ReleasedF arxiv.org/abs/2610.10270 · Retrieved 2 November 2026Note. [A] “Submitted on 7 Oct 2026”
Input and output3/4 disclosed
Model typeWorld action model
Source and note: Model typeD github.com/roboterax/video-prediction-policy-2 · Retrieved 2 November 2026Note. RobotEra's own term in the code repository: “a world-action model (WAM)”.
InputsCamera Language Robot state
Source and note: InputsA github.com/roboterax/video-prediction-policy-2/blob/main/docs/robodojo.md · Retrieved 2 November 2026Note. RoboDojo guide: “The policy reads RGB, robot state and instructions through the standard interface.”
What it controlsArms and hands
Source and note: What it controlsA github.com/roboterax/video-prediction-policy-2/blob/main/docs/robodojo.md · Retrieved 2 November 2026Note. RoboDojo guide, action format of the dual-arm setup: “State/action is dual-arm absolute EE16”.
Degrees of freedom controlledNot stated
Source and note: Degrees of freedom controlledA github.com/roboterax/video-prediction-policy-2/blob/main/docs/robodojo.md · Retrieved 2 November 2026Note. The guide gives an action-vector format for a benchmark setup, not a count of controlled joints: “State/action is dual-arm absolute EE16”.
Size and speed0/2 disclosed
ParametersNot stated
Source and note: ParametersB arxiv.org/html/2610.10270 · Retrieved 2 November 2026Note. The paper gives no total count; it names its video backbone and its action expert: “VPP2 builds upon the pretrained Wan2.1-I2V-14B model”; “The action expert is a 0.9B-parameter diffusion transformer (DiT)”.
Control rateNot stated
Source and note: Control rateB arxiv.org/html/2610.10270 · Retrieved 2 November 2026Note. The paper gives a latency, not a control rate: “The total latency per action chunk is therefore approximately 0.22 seconds.”
Where it runs and what it is built on1/2 disclosed
On robot or cloudNot stated
Base modelWan2.1-I2V-14B
Source and note: Base modelB arxiv.org/html/2610.10270 · Retrieved 2 November 2026Note. Paper: “VPP2 builds upon the pretrained Wan2.1-I2V-14B model”.
Openness1/3 disclosed
AccessOpen weights (download)
Source and note: AccessC huggingface.co/Haodong082399/VPP2 · Retrieved 2 November 2026Note. Hugging Face model page, listed as not gated: “Official VPP2 weights for zero-shot video prediction, RoboDojo and LIBERO.”
Licence documentNot stated
Source and note: Licence documentC huggingface.co/Haodong082399/VPP2 · Retrieved 2 November 2026Note. The code repository carries the MIT License, but its notices say: “Model weights and datasets are distributed separately under their own terms;” The Hugging Face page names no licence for the VPP2 weights.
Licence typeNot stated
Source and note: Licence typeC huggingface.co/Haodong082399/VPP2 · Retrieved 2 November 2026Note. The code repository carries the MIT License, but its notices say: “Model weights and datasets are distributed separately under their own terms;” The Hugging Face page names no licence for the VPP2 weights.
The developer’s own claims1/2 disclosed
Learns new tasksWithout examples
Source and note: Learns new tasksE robert-gyj.github.io/video-prediction-policy-2 · Retrieved 2 November 2026Note. Developer claim: “We introduce Video Prediction Policy 2 (VPP2), a WAM that enables strong zero-shot generalization in both video prediction and action generation.” The paper's robot test: “VPP2 exhibits strong zero-shot manipulation capabilities on a real-world ALOHA platform”. The benchmark results use “following benchmark-specific post-training”.
Works across robotsNot stated
Source and note: Works across robotsB arxiv.org/html/2610.10270 · Retrieved 2 November 2026Note. The paper's real-robot test names one platform: “VPP2 exhibits strong zero-shot manipulation capabilities on a real-world ALOHA platform”.Yes 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
VPP2 is a joint work of RobotEra with Tsinghua University, HKUST (GZ), UC Berkeley and Shanghai Jiao Tong University.
The Hugging Face page also lists video-only checkpoints for zero-shot video prediction.
Sources
One line per source, not per figure: a source has one retrieval date.
- Agithub.com/roboterax/video-prediction-policy-2/blob/main/docs/robodojo.mdRetrieved 2 November 2026
- Barxiv.org/html/2610.10270Retrieved 2 November 2026
- Chuggingface.co/Haodong082399/VPP2Retrieved 2 November 2026
- Dgithub.com/roboterax/video-prediction-policy-2Retrieved 2 November 2026
- Erobert-gyj.github.io/video-prediction-policy-2Main source. Retrieved 2 November 2026
- Farxiv.org/abs/2610.10270Retrieved 2 November 2026