What’s Out There
← Back to swiping
Images & videoFreeCompany

AgiBot World 2026

AgiBot

Real-world robot demonstrations from homes and commercial spaces, collected on the AgiBot G2 platform.

A dual-arm robot manipulation dataset collected entirely in real environments, in LeRobot format. One row is an episode of robot camera video, joint states, and actions with annotations. It targets imitation learning for embodied AI.

Access

Cost
FreeNo charge; very large downloads.
License
Open, non-commercial onlyCC BY-NC-SA 4.0
Sign-up
No sign-up
How you get it
API and Bulk downloadLeRobot (Parquet and MP4)

What's in it

One row is
One robot episode
Covers
Commercial and home environments (location not stated)
Years
2026
Updated
Irregular
Size
Between 1,000 and 10,000 episodes per the card; exact total not stated
camera videojoint statesactionstask instructionscene type

Quality

3 of 5

Real-world data from one robot platform with LeRobot compatibility; non-commercial terms and a single embodiment limit transfer.

  • One robot design
  • Non-commercial license
  • Annotation depth not detailed in the record

Connects to

  • ABC-130k: Large bimanual teleoperation set on a different robot
  • FineART: Bimanual dataset with dense subtask language, in the same LeRobot format

Questions it could answer

  • How are episodes distributed across scene types?
  • How long do tasks run, and where do they fail?

Similar data sets

  • FineART

    40,543 robot episodes and 533,913 human-written subtask labels covering 1,718 hours.

    Images & videoScale AI and Hugging Face LeRobotFree

  • ABC-130k

    Billed as the largest open-source robot teleoperation dataset: bimanual trajectories on two-arm YAM stations.

    Images & videoXDOFFree

  • OpenRoboCare

    315 sessions of occupational therapists doing caregiving tasks, with video, touch, and gaze.

    Images & videoEmPRISE Lab at Cornell UniversityFree

  • Ego4D

    3,670 hours of daily life filmed from head-mounted cameras by 923 people in 9 countries.

    Images & videoEgo4D consortium (Meta AI and university partners)By application

  • Kinetics-700

    650,000 ten-second YouTube clips of people doing 700 things, from yoga to changing a tire.

    Images & videoGoogle DeepMind (hosted by the Common Visual Data Foundation)Free

  • GDPval

    220 real work tasks across 44 occupations, with reference files, to test AI on economically valuable output.

    LaborOpenAIFree