Skip to content
embodied.guide

Amap CV LabChina2026

ABot-M0

Key figures as stated by the developer

6 of 7 key figures stated
  • Model type ⁠C vision-language-action
  • Controls ⁠A arms and hands
  • Inputs ⁠A camera language robot state
  • Access ⁠B Open weights (download)
  • Base model ⁠A Qwen3-VL-4B
  • Licence ⁠B Free, commercial use

Not stated: Parameters

Full specifications

8 of 14 fields stated
The letter points to the source at the foot of the page. Fields without a figure are dashed: the developer does not state them — that is not the same as zero, and not the same as no.
Field Value Source and caveats
Release
Released11 February 2026

[A] "Date: February 11, 2026" ⁠A
Input and output
Model typevision-language-action

Paper title: "ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learning". ⁠C

Inputscamera language robot state

Technical report: "The VLM processes stacked multi-view image sequences, typically captured from front-facing, wrist-mounted, and top-down cameras, together with natural language instructions." The action expert also takes "the current robot state". ⁠A

Controlsarms and hands

Technical report: "The model always generates full dual-arm outputs in length of 14". ⁠A

Degrees of freedom controllednot stated
Size and speed
Parametersnot stated

Note The report names the parts, not a total: "ABot-M0 employs Qwen3-VL [ 1 ] 4B as its VLM backbone, the 0.16B DiT with Action Manifold Learning (AML) as the action expert". ⁠A

Control ratenot stated
Where it runs and what it is built on
On robot or cloudnot stated
Base modelQwen3-VL-4B

Technical report: "ABot-M0 employs Qwen3-VL [ 1 ] 4B as its VLM backbone, the 0.16B DiT with Action Manifold Learning (AML) as the action expert". ⁠A
Openness
AccessOpen weights (download)

Model card for one released checkpoint: "This is the model weight for Libero Benchmark of our proposed model" ABot-M0. ⁠B

Licence termsApache-2.0

Model card header: "license: apache-2.0". ⁠B

LicenceFree, commercial use

Model card header: "license: apache-2.0". ⁠B
The developer’s own claims
Learns new tasksnot stated
Works across robotsnot stated

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.
How to read the figures
50HzA
The letter in the grey square after a figure points to the source the figure came from. Pages with several sources carry several letters.

What the data labels mean

Notes

  • The ABot-M0 paper gives an Alibaba email address for its corresponding author.
Versions and notes in the key figures (7)

Model type

Paper title: "ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learning".

Controls

Technical report: "The model always generates full dual-arm outputs in length of 14".

Inputs

Technical report: "The VLM processes stacked multi-view image sequences, typically captured from front-facing, wrist-mounted, and top-down cameras, together with natural language instructions." The action expert also takes "the current robot state".

Access

Model card for one released checkpoint: "This is the model weight for Libero Benchmark of our proposed model" ABot-M0.

Parameters

Note The report names the parts, not a total: "ABot-M0 employs Qwen3-VL [ 1 ] 4B as its VLM backbone, the 0.16B DiT with Action Manifold Learning (AML) as the action expert".

Base model

Technical report: "ABot-M0 employs Qwen3-VL [ 1 ] 4B as its VLM backbone, the 0.16B DiT with Action Manifold Learning (AML) as the action expert".

Licence

Model card header: "license: apache-2.0".

Sources

One line per source, not per figure: a source has one retrieval date. The letter beside a figure points to the line it came from.