# octo-models/octo

Octo is a transformer-based robot policy trained on a diverse mix of 800k robot trajectories.

Repository: https://github.com/octo-models/octo
Canonical: https://ross.abutalabs.com/products/octo
Homepage: https://octo-models.github.io/
Language: Python
License: MIT
License Family: permissive
Last push: 2024-07-31T00:26:15+00:00

## Health v2 (maintenance only)
Score: 17/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 71
- inputs: {"age_days": 994, "days_push": 764, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1751, forks 281 (observed 2026-08-28T04:05:31.150809+00:00)

## What it is
Octo is an open-source generalist robot policy: a transformer-based diffusion policy pretrained on 800k robot trajectories from the Open X-Embodiment dataset. It provides pretrained checkpoints (Octo-Small and Octo-Base) plus training and finetuning code built on JAX for adapting the policy to new robots, sensors, and action spaces.

## Use cases
- control a robot arm with a pretrained generalist policy
- finetune a robot policy to my custom robot setup with a small dataset
- run zero-shot evaluation of a robot manipulation model
- instruct a robot via language commands or goal images
- research generalist robot policies and multi-embodiment learning
- load a pretrained robot policy checkpoint from Hugging Face

## When to choose
- you need a pretrained generalist manipulation policy for robot arms
- you want to finetune a robot foundation model with limited compute
- you work with RGB camera inputs, language instructions, or goal-image conditioning
- you are doing robotics research on multi-task, multi-embodiment policies

## When to avoid
- you need locomotion, navigation, or mobile robot policies rather than manipulation
- you require real-time control on low-power hardware without a GPU
- you want a plug-and-play production robotics stack rather than a research codebase
- you need non-Python or non-JAX integration

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference
- domain: robotics, machine-learning, deep-learning, artificial-intelligence
- platform: python, cross-platform
- tags: robot-policy, diffusion-policy, transformer, jax, open-x-embodiment, robot-manipulation, finetuning, generalist-robot-policy, gpu, linux

## Member repositories
- octo-models/octo (main) score 17

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:31.150809+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 2026-08-30T03:28:33.721452+00:00, confidence not recorded.
  - readme: https://github.com/octo-models/octo (fetched 2026-08-28T04:05:31.150809+00:00, sha 2e5d00eea91f)
  - homepage: https://octo-models.github.io/ (fetched 2026-08-29T11:06:24.392503+00:00, sha fb16cb6b6fc9)
- Data as of 2026-08-30T08:39:29.467469+00:00.