Amap CV Lab · robot AI model
ABot-ER
China · 2026
Runs on R1 Pro by Galaxea DynamicsSource and note: R1 ProA amap-cvlab.github.io/ABot-ER/ · Retrieved 9 November 2026
Sources retrieved: 9 November 2026 (A, B, C, D, E)
- Model type
- Embodied reasoningSource and note: Model typeC arxiv.org/abs/2610.18355 · Retrieved 9 November 2026Note. AMAP CV Lab’s own term: “ABot Embodied Reasoning (ABot-ER), a unified embodied reasoning foundation model”.
- Inputs
- Camera LanguageSource and note: InputsA amap-cvlab.github.io/ABot-ER/ · Retrieved 9 November 2026Note. Project page: “Images, video frames, instructions, and available state or history are encoded into one shared token sequence.”
- Access
- Open weights (download)Source and note: AccessB github.com/amap-cvlab/ABot-ER · Retrieved 9 November 2026Note. README: “The table lists the released models for embodied QA and task planning.” and “ABot-ER model pages and training data are now public on Hugging Face. The 35B-A3B page links to ModelScope for weight downloads.”
Specifications
4 of 14 stated 10 not stated
5 fields hidden 5 not-stated fields with a note stay ·
How to read the figuresShow guideClose
- not statedNo 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/2610.18355 · Retrieved 9 November 2026Note. [A] The arXiv listing for the technical report: “Submitted on 13 Oct 2026”.
Input and output2/4 disclosed
Model typeEmbodied reasoning
Source and note: Model typeC arxiv.org/abs/2610.18355 · Retrieved 9 November 2026Note. AMAP CV Lab’s own term: “ABot Embodied Reasoning (ABot-ER), a unified embodied reasoning foundation model”.
InputsCamera Language
Source and note: InputsA amap-cvlab.github.io/ABot-ER/ · Retrieved 9 November 2026Note. Project page: “Images, video frames, instructions, and available state or history are encoded into one shared token sequence.”
What it controlsNot stated
Source and note: What it controlsA amap-cvlab.github.io/ABot-ER/ · Retrieved 9 November 2026Note. Project page: “ABot-ER predicts assessments, plans, and interaction coordinates—not motor commands.”
Degrees of freedom controlledNot stated
Size and speed0/2 disclosed
ParametersNot stated
Source and note: ParametersB github.com/amap-cvlab/ABot-ER · Retrieved 9 November 2026Note. Released in five sizes: ABot-ER-2B, ABot-ER-4B, ABot-ER-9B, ABot-ER-27B and ABot-ER-35B-A3B.
Control rateNot stated
Where it runs and what it is built on0/2 disclosed
On robot or cloudNot stated
Base modelNot stated
Source and note: Base modelE huggingface.co/acvlab/ABot-ER-9B · Retrieved 9 November 2026Note. The Hugging Face model cards list a different backbone per size: “Qwen3.5-2B” (2B), “Qwen3.5-4B” (4B), “Qwen3.5-9B” (9B) and “Qwen3.8-27B” (27B).
Openness1/3 disclosed
AccessOpen weights (download)
Source and note: AccessB github.com/amap-cvlab/ABot-ER · Retrieved 9 November 2026Note. README: “The table lists the released models for embodied QA and task planning.” and “ABot-ER model pages and training data are now public on Hugging Face. The 35B-A3B page links to ModelScope for weight downloads.”
Licence documentNot stated
Source and note: Licence documentD huggingface.co/acvlab/ABot-ER-9B/blob/main/LICENSE · Retrieved 9 November 2026Note. The Hugging Face model card lists the licence as “aboter-backbone-and-component-terms”; its LICENSE file says: “No new model-weight license is selected by this packaging update.” The README says: “Model weights, datasets, simulator assets, and third-party code retain their respective licenses.” The project code is Apache-2.0.
Licence typeNot stated
The developer’s own claims0/2 disclosed
Learns new tasksNot stated
Works across robotsNot stated
Source and note: Works across robotsYes 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
AMAP CV Lab, Alibaba Group, says it developed the ABot-ER model family.
AMAP CV Lab says the model generates answers, coordinates and plans, and that an agent runtime or downstream controller executes the actions.
AMAP CV Lab says the model is intended for research and evaluation.
Sources
One line per source, not per figure: a source has one retrieval date.
- Aamap-cvlab.github.io/ABot-ER/Main source. Retrieved 9 November 2026
- Bgithub.com/amap-cvlab/ABot-ERRetrieved 9 November 2026
- Carxiv.org/abs/2610.18355Retrieved 9 November 2026
- Dhuggingface.co/acvlab/ABot-ER-9B/blob/main/LICENSERetrieved 9 November 2026
- Ehuggingface.co/acvlab/ABot-ER-9BRetrieved 9 November 2026
