Sber Robotics Center · robot AI model
GreenWAM
Russia · 2026
Runs on Green by SberSource and note: GreenA greenvla.github.io/GreenWAM/ · Retrieved 19 November 2026
Sources retrieved: 19 November 2026 (A, B, C)
- Model type
- World modelSource and note: Model typeC arxiv.org/abs/2610.18774 · Retrieved 19 November 2026Note. “We present GreenWAM, a world foundation model for robotics, together with the recipe that produces it”. Also serves as a policy in world-action mode.
- What it controls
- Arms and handsSource and note: What it controlsB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. Covers “20 robot families, from bimanual arms to dexterous humanoids, are described in common joint coordinates” (79 shared anatomical slots).
- Inputs
- Camera Language Robot stateSource and note: InputsA greenvla.github.io/GreenWAM/ · Retrieved 19 November 2026Note. In world-action mode, “the instruction, the multi-camera canvas and the joint state enter a frozen reasoner, a trained generator for the video tokens and, in the adapted policy, an action expert.” “The checkpoints also need the Wan2.2 VAE and the Cosmos 3 Edge text tokenizer at the revision they were evaluated with:”
- Access
- Open weights (download)Source and note: AccessB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. “This repository holds the four benchmark checkpoints” of the technical report.
- Base model
- Cosmos 3 EdgeSource and note: Base modelB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. GreenWAM “is one robot world model that serves both roles across different bodies. It is built on Cosmos 3 Edge”.
- Licence document
- OpenMDW-1.1Source and note: Licence documentB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. Weights are released under OpenMDW License 1.1, the license of Cosmos 3 Edge.

Specifications
9 of 14 stated 5 not stated
3 fields hidden 2 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
Released13 October 2026
Source and note: ReleasedC arxiv.org/abs/2610.18774 · Retrieved 19 November 2026Note. arXiv submission date: “Submitted on 13 Oct 2026”.
Input and output3/4 disclosed
Model typeWorld model
Source and note: Model typeC arxiv.org/abs/2610.18774 · Retrieved 19 November 2026Note. “We present GreenWAM, a world foundation model for robotics, together with the recipe that produces it”. Also serves as a policy in world-action mode.
InputsCamera Language Robot state
Source and note: InputsA greenvla.github.io/GreenWAM/ · Retrieved 19 November 2026Note. In world-action mode, “the instruction, the multi-camera canvas and the joint state enter a frozen reasoner, a trained generator for the video tokens and, in the adapted policy, an action expert.” “The checkpoints also need the Wan2.2 VAE and the Cosmos 3 Edge text tokenizer at the revision they were evaluated with:”
What it controlsArms and hands
Source and note: What it controlsB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. Covers “20 robot families, from bimanual arms to dexterous humanoids, are described in common joint coordinates” (79 shared anatomical slots).
Degrees of freedom controlledNot stated
Source and note: Degrees of freedom controlledB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. Uses a shared anatomical action space of 79 slots where “20 robot families, from bimanual arms to dexterous humanoids, are described in common joint coordinates”.
Size and speed1/2 disclosed
ParametersNot stated
Control rate28.3Hz
Source and note: Control rateA greenvla.github.io/GreenWAM/ · Retrieved 19 November 2026Note. “The policy is distilled into a one-step student that answers in 35.4 ms on one RTX 4090, at 28.3 Hz.”
Where it runs and what it is built on1/2 disclosed
On robot or cloudNot stated
Base modelCosmos 3 Edge
Source and note: Base modelB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. GreenWAM “is one robot world model that serves both roles across different bodies. It is built on Cosmos 3 Edge”.
Openness2/3 disclosed
AccessOpen weights (download)
Source and note: AccessB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. “This repository holds the four benchmark checkpoints” of the technical report.
Licence documentOpenMDW-1.1
Source and note: Licence documentB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. Weights are released under OpenMDW License 1.1, the license of Cosmos 3 Edge.
Licence typeNot stated
Source and note: Licence typeB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. Weights are released under OpenMDW License 1.1, the license of Cosmos 3 Edge.
The developer’s own claims1/2 disclosed
Learns new tasksNot stated
Works across robotsYes
Source and note: Works across robotsB huggingface.co/SberRoboticsCenter/GreenWAM · Retrieved 19 November 2026Note. GreenWAM “is one robot world model that serves both roles across different bodies. It is built on Cosmos 3 Edge”. Covers “20 robot families, from bimanual arms to dexterous humanoids, are described in common joint coordinates”.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
Described by Sber Robotics Center as “We present GreenWAM, a world foundation model for robotics, together with the recipe that produces it”.
Four benchmark checkpoints and evaluation code are publicly released.
Sources
One line per source, not per figure: a source has one retrieval date.
- Agreenvla.github.io/GreenWAM/Main source. Retrieved 19 November 2026
- Bhuggingface.co/SberRoboticsCenter/GreenWAMRetrieved 19 November 2026
- Carxiv.org/abs/2610.18774Retrieved 19 November 2026

