National University of Singapore · robot AI model
OpenWAM-α
Singapore · 2026
Sources retrieved: 2 November 2026 (A, B, C, D)
- Model type
- World action modelSource and note: Model typeB github.com/OpenWAM-Official/OpenWAM · Retrieved 2 November 2026Note. Developer wording: “An open pretrained WAM that applies these principles at scale”.
- What it controls
- Arms and handsSource and note: What it controlsA arxiv.org/pdf/2609.07398 · Retrieved 2 November 2026Note. The action vector: “an 80-dimensional vector with fixed slot semantics, comprising two mirrored 34-D arm blocks”, each with end-effector pose, gripper and a dexterous hand.
- Inputs
- Camera Language Robot stateSource and note: InputsA arxiv.org/pdf/2609.07398 · Retrieved 2 November 2026Note. Technical report: “all camera views tiled into one canvas”; “the robot state at the start of the window supplies the proprioception token”. Language: “Maps the instruction into a 4096-dim context that both streams consume through their own cross-attention”.
- Access
- Open weights (download)Source and note: AccessB github.com/OpenWAM-Official/OpenWAM · Retrieved 2 November 2026Note. Developer wording: “[2026/09/06] 🤗 We release all pretrained and finetuned models on huggingface”.
- Base model
- Wan2.2-TI2V-5BSource and note: Base modelC huggingface.co/OpenWAM/OpenWAM-Alpha-Pretrain-Foundation-Model · Retrieved 2 November 2026Note. Model card: “a Wan2.2-TI2V-5B video backbone with its native Wan-VAE encoder”.
- Licence type
- Free, commercial useSource and note: Licence typeD openwam-official.github.io · Retrieved 2 November 2026Note. Developer wording: “We release the full stack: infrastructure, evaluation protocols, pretrained models, and data recipes”.
Specifications
10/14 specifications disclosed 4 not stated
2 fields hidden 2 not-stated fields with a note stay ·
How to read the figuresShow guideClose
Not statedThe developer publishes nothing on itYesThe developer says it does
Select a source marker for source details and notes.
Release1/1 disclosed
Released6 September 2026
Source and note: ReleasedB github.com/OpenWAM-Official/OpenWAM · Retrieved 2 November 2026Note. [A] “[2026/09/06] 🤗 We release all pretrained and finetuned models on huggingface”
Input and output3/4 disclosed
Model typeWorld action model
Source and note: Model typeB github.com/OpenWAM-Official/OpenWAM · Retrieved 2 November 2026Note. Developer wording: “An open pretrained WAM that applies these principles at scale”.
InputsCamera Language Robot state
Source and note: InputsA arxiv.org/pdf/2609.07398 · Retrieved 2 November 2026Note. Technical report: “all camera views tiled into one canvas”; “the robot state at the start of the window supplies the proprioception token”. Language: “Maps the instruction into a 4096-dim context that both streams consume through their own cross-attention”.
What it controlsArms and hands
Source and note: What it controlsA arxiv.org/pdf/2609.07398 · Retrieved 2 November 2026Note. The action vector: “an 80-dimensional vector with fixed slot semantics, comprising two mirrored 34-D arm blocks”, each with end-effector pose, gripper and a dexterous hand.
Degrees of freedom controlledNot stated
Source and note: Degrees of freedom controlledA arxiv.org/pdf/2609.07398 · Retrieved 2 November 2026Note. The size of the shared action vector, not a figure for a named robot: “an 80-dimensional vector with fixed slot semantics, comprising two mirrored 34-D arm blocks”.
Size and speed0/2 disclosed
ParametersNot stated
Source and note: ParametersA arxiv.org/pdf/2609.07398 · Retrieved 2 November 2026Note. The report gives no total count; it names a “a dedicated 1B-parameter ActionDiT” next to the Wan2.2-TI2V-5B video backbone.
Control rateNot stated
Where it runs and what it is built on1/2 disclosed
On robot or cloudNot stated
Base modelWan2.2-TI2V-5B
Source and note: Base modelC huggingface.co/OpenWAM/OpenWAM-Alpha-Pretrain-Foundation-Model · Retrieved 2 November 2026Note. Model card: “a Wan2.2-TI2V-5B video backbone with its native Wan-VAE encoder”.
Openness3/3 disclosed
AccessOpen weights (download)
Source and note: AccessB github.com/OpenWAM-Official/OpenWAM · Retrieved 2 November 2026Note. Developer wording: “[2026/09/06] 🤗 We release all pretrained and finetuned models on huggingface”.
Licence documentApache-2.0
Source and note: Licence documentC huggingface.co/OpenWAM/OpenWAM-Alpha-Pretrain-Foundation-Model · Retrieved 2 November 2026Note. Model card header: “license: apache-2.0”. The code repository: “OpenWAM is released under the Apache License 2.0”.
Licence typeFree, commercial use
Source and note: Licence typeD openwam-official.github.io · Retrieved 2 November 2026Note. Developer wording: “We release the full stack: infrastructure, evaluation protocols, pretrained models, and data recipes”.
The developer’s own claims2/2 disclosed
Learns new tasksNeeds retraining
Source and note: Learns new tasksB github.com/OpenWAM-Official/OpenWAM · Retrieved 2 November 2026Note. Developer wording: “For fine-tuning, set training.finetune_ckpt_path”. Model card: “This is the pretrained foundation model itself, before any downstream finetuning.”
Works across robotsYes
Source and note: Works across robotsD openwam-official.github.io · Retrieved 2 November 2026Note. Developer claim: the simulation and real-robot experiments “which together span embodiments from single-arm and bimanual manipulation to dexterous hands”.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
OpenWAM-α is the pretrained model of the OpenWAM research stack, written by authors from seven universities, the first of them the National University of Singapore.
The OpenWAM project released its code and models on 6 September 2026.
The OpenWAM Hugging Face organisation lists 46 models: the pretrained foundation model and checkpoints fine-tuned for single benchmarks and robots.
Sources
One line per source, not per figure: a source has one retrieval date.
- Aarxiv.org/pdf/2609.07398Retrieved 2 November 2026
- Bgithub.com/OpenWAM-Official/OpenWAMRetrieved 2 November 2026
- Chuggingface.co/OpenWAM/OpenWAM-Alpha-Pretrain-Foundation-ModelRetrieved 2 November 2026
- Dopenwam-official.github.ioMain source. Retrieved 2 November 2026