← Sinan Lab › All datasets › AgiBot DigitalWorld (+ AgiBot World Challenge datasets)
Dataset · AgiBot (AgiBot World / OpenDriveLab)

AgiBot DigitalWorld (+ AgiBot World Challenge datasets)

Landing page →Hugging Face →China mirror →

Simulation (digital twins of AgiBot World tasks); scale stated in steps not episodes. Challenge sets on HF (all cc-by-nc-sa-4.0): AgiBotWorldChallenge-2025 (3.1 TB; Manipulation track 10 tasks with 'hundreds of trajectories per task' sim+real via Genie Sim; World Model track 'over 30,000' training trajectories; v1.0.0 25-06-2025, v1.1.0 11-08-2025), AgiBotWorldChallenge-202509, AgiBotWorldChallenge-2026. Real AgiBot World Alpha 92,214 / Beta 1,003,672 trajectories in LeRobot format (GitHub README).

Facts

Organisation
AgiBot (AgiBot World / OpenDriveLab)
Form tags
dual arm · simulation
Embodiments (as stated)
AgiBot dual-arm robot, 34 DoF (2 DoF head, 2 DoF waist, 7 DoF per arm, 8 DoF per gripper)
Episodes
unverified
Hours
unverified
Frames
unverified
Tasks
long-horizon digital-twin tasks + agent-generated atomic tasks (count not stated); 180+ object classes
Scenes
5
Modalities
rgb · depth · state
Format
custom (MP4 RGB + PNG depth + HDF5 proprioception + JSON task info); conversion scripts to LeRobot v2.0 provided
Licence
CC BY-NC-SA 4.0
Commercial use
not allowed (stated)
Released
2025-02
Size
7610 GB
Sources · 9 sources
embodiments https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld/raw/main/README.md Official page · Fetched 2026-09-18 · No snapshot stored yet Dual-arm robot with 34 DoF (2 head, 2 waist, 7 per arm, 8 per gripper)
tasks https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld/raw/main/README.md Official page · Fetched 2026-09-18 · No snapshot stored yet '1 million+ steps of enhanced digital twins of long-horizon real-world tasks' and '500,000+ steps of atomic tasks automatically generated by agents'; '180+ classes of objects'
scenes https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld/raw/main/README.md Official page · Fetched 2026-09-18 · No snapshot stored yet '5 classes of scenes'
modalities https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld/raw/main/README.md Official page · Fetched 2026-09-18 · No snapshot stored yet RGB video, depth (PNG), proprioceptive state (joint positions/velocities, end-effector pose, gripper, odometry)
format https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld/raw/main/README.md Official page · Fetched 2026-09-18 · No snapshot stored yet HDF5 proprio + JSON task info + video/depth folders; LeRobot v2.0 conversion scripts
license https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld Official page · Fetched 2026-09-18 · No snapshot stored yet 'CC BY-NC-SA 4.0' (README); GitHub README also states CC BY-NC-SA 4.0; ModelScope API metadata lists 'Apache License 2.0'
mirror_url https://modelscope.cn/api/v1/datasets/agibot_world/AgiBotDigitalWorld Official page · Fetched 2026-09-18 · No snapshot stored yet ModelScope API confirms Name 'AgiBotDigitalWorld', Namespace 'agibot_world', storage ~576.6 GB
release_date https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld Official page · Fetched 2026-09-18 · No snapshot stored yet Released on Hugging Face '2025/2/24'
size_gb https://huggingface.co/datasets/agibot-world/AgiBotDigitalWorld Official page · Fetched 2026-09-18 · No snapshot stored yet HF card '7.61 TB'

Related bodies in the index

Matched by stated robot names or form; a match means the form is covered, not that a model was trained on it.

SinanLab · Sinan LabClear direction, grounded judgement.