Shanghai Innovation Institute · robot AI model
τ₀-WM
China · 2026
Sources retrieved: 19 November 2026 (A, B, C)
- Model type
- World action modelSource and note: Model typeA huggingface.co/sii-research/tau-0-wm/raw/main/README.md · Retrieved 19 November 2026Note. Model card: “τ₀-World Model (τ₀-WM) is a unified video-action world model that integrates policy learning, video prediction, and action evaluation within a single future-predictive framework.”
- What it controls
- Arms and handsSource and note: What it controlsA huggingface.co/sii-research/tau-0-wm/raw/main/README.md · Retrieved 19 November 2026Note. Model card: “The model is designed for dual-arm manipulation with the following specifications:” “Absolute poses of end-effectors and gripper openness (16 channels total).”
- Inputs
- Camera Language Robot stateSource and note: InputsA huggingface.co/sii-research/tau-0-wm/raw/main/README.md · Retrieved 19 November 2026Note. Model card: “Jointly predicts future visual latents and continuous action chunks from multi-view observations, language instructions, and robot state.”
- Access
- Open weights (download)Source and note: AccessA huggingface.co/sii-research/tau-0-wm/raw/main/README.md · Retrieved 19 November 2026Note. Checkpoints for both the Video Action Model and simulator are hosted on Hugging Face: “hf download sii-research/tau-0-wm”
- Parameters
- 5.5BSource and note: ParametersB arxiv.org/pdf/2606.01027 · Retrieved 19 November 2026Note. Paper: “Together, they form a 5.5B-parameter Video Action Model.” (5B video transformer plus 0.5B action decoder).
- Base model
- Wan2.2-TI2V-5BSource and note: Base modelB arxiv.org/pdf/2606.01027 · Retrieved 19 November 2026Note. Paper: “VAM is instantiated from Wan2.2-TI2V-5B”
- Licence type
- Free, commercial useSource and note: Licence typeA huggingface.co/sii-research/tau-0-wm/raw/main/README.md · Retrieved 19 November 2026Note. Model card metadata: “license: apache-2.0”

Specifications
11 of 14 stated 3 not stated
0 fields hidden 3 not-stated fields with a note stay ·
How to read the figuresShow guideClose
- not statedNo source
- yesStated by a source
- 0Reported as zero
- ASource
Select a source marker for source details and notes.
Release1/1 disclosed
Input and output3/4 disclosed
Size and speed1/2 disclosed
Where it runs and what it is built on1/2 disclosed
Openness3/3 disclosed
The developer’s own claims2/2 disclosed
Notes
Developed by Shanghai Innovation Institute and AGIBOT Finch. “τ₀-World Model (τ₀-WM) is a unified video-action world model that integrates policy learning, video prediction, and action evaluation within a single future-predictive framework.”
The arXiv preprint was submitted on 31 May 2026, and checkpoints are available for open download on Hugging Face.
Sources
One line per source, not per figure: a source has one retrieval date.
- Ahuggingface.co/sii-research/tau-0-wm/raw/main/README.mdMain source. Retrieved 19 November 2026
- Barxiv.org/pdf/2606.01027Retrieved 19 November 2026
- Carxiv.org/abs/2606.01027Retrieved 19 November 2026




