# OpenDriveLab/AgiBot-World

[IROS 2025 Best Paper Award Finalist & IEEE TRO 2026] The Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems

Repository: https://github.com/OpenDriveLab/AgiBot-World
Canonical: https://ross.abutalabs.com/products/agibot-world
Homepage: https://opendrivelab.com/AgiBot-World/
Language: Python
License Family: other
Topics: robotic-manipulation, vision-language-action-model, pretraining-for-robotics, robotic-foundation-model
Last push: 2026-05-29T12:31:22+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 84, release rhythm 35, longevity 45
- inputs: {"age_days": 633, "days_push": 96, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3157, forks 217 (observed 2026-08-28T04:07:46.434445+00:00)

## What it is
AgiBot World Colosseo is a large-scale open-source robot manipulation dataset with over one million trajectories across 217 real-world tasks, collected by 100+ homogeneous robots, plus the GO-1 robotic foundation model pretrained on it. It provides a full-stack embodied AI ecosystem including benchmarks, task catalogs, and lightweight model variants for scalable robot learning.

## Use cases
- train a robot manipulation policy on large-scale real-world data
- pretrain a vision-language-action model for bimanual manipulation
- benchmark generalist robot policies against Open X-Embodiment baselines
- download robot trajectories for dexterous and long-horizon manipulation research
- fine-tune the GO-1 foundation model on custom robot tasks
- find embodied AI datasets for dual-arm and collaborative manipulation

## When to choose
- you need large-scale, high-fidelity real-world manipulation data for robot learning
- you want a pretrained generalist policy baseline for manipulation research
- your work targets bimanual, dexterous, or long-horizon real-world tasks

## When to avoid
- you need a non-commercial license dataset (CC BY-NC-SA 4.0)
- your focus is simulation-only or low-cost tabletop lab data
- you lack storage/compute for multi-terabyte datasets

## Facets
- artifact type: dataset
- maturity: active
- function: machine-learning, deep-learning, robotics, data-science
- domain: robotics, machine-learning, artificial-intelligence
- platform: python
- tags: robot-manipulation, embodied-ai, vision-language-action, robot-foundation-model, bimanual-manipulation, imitation-learning, huggingface-dataset, go-1, datasets, gpu, linux

## Member repositories
- OpenDriveLab/AgiBot-World (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:46.434445+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:25:36.779809+00:00, confidence not recorded.
  - readme: https://github.com/OpenDriveLab/AgiBot-World (fetched 2026-08-28T04:07:46.434445+00:00, sha b556b1929b23)
  - homepage: https://opendrivelab.com/AgiBot-World/ (fetched 2026-08-29T09:40:02.678543+00:00, sha a98eb2d6a267)
- Data as of 2026-08-30T08:39:29.467469+00:00.