Octo
SoftwareTransformer diffusion robot policies with pretrained checkpoints and fine-tuning tools; observation history, action statistics and robot-specific interfaces must match the selected model. [1]
Installation & first result →Official websiteGetting started
- For
- Robot-learning researchers with a compatible environment and accelerator setup [1]
- Source rights
- Some licenses are unresolved. Review licenses
- Unknowns
- operating system, compute requirements not fully established. Review gaps
Source researched · Not tested by us
Sources checked · Research details
Licenses & related hardware
- License
- MIT · license not fully assessed for datasets, documentation · Details by part
Overview
Octo is a family of transformer diffusion policies with code for training, fine-tuning and evaluation. The reviewed source is v1.5. This entry also inspects the separately hosted Octo Base 1.5 model card, which describes a 93-million-parameter model trained on a mixture of Open X-Embodiment datasets. Code and checkpoint revisions are separate parts of the evidence. [1] [2] [8]
Is this for you?
- Workflows
- Load pretrained checkpoints
- Run offline inference with matched dataset statistics
- Fine-tune for new observations and action spaces
- Evaluate through an environment adapter
- Known limitations
- Action chunks are not complete task trajectories
- Missing modalities and absent history require appropriate masks
- Pretraining dataset and compute are substantially larger than the inference example
- Model and source licenses do not establish a uniform license for every training dataset
Last item is an editorial rights-scope limitation.
What you need
- Languages / runtimes
- Python
Installation & first result
Open the upstream getting-started guide [5] · Project documentation [1] [4]
Hardware, models & data
Project notes
Select a versioned checkpoint
The source overview's quick-start example loads octo-base-1.5, while its checkpoint table links to unversioned Base and Small model names. The inference notebook uses Small 1.5. Record the exact model repository and revision used in an experiment rather than relying on the family name. The model card inspected here is for Base 1.5; its dimensions should not be assumed to describe a subsequently fine-tuned model. [1] [5] [8]
Upstream's installation route uses Python 3.10 and JAX 0.4.20, with separate CUDA11 and TPU commands. Its initial debug fine-tuning command loads Small 1.5. Those pinned instructions need a compatible accelerator environment; this review did not install or execute them. The reported pretraining route is much larger: upstream describes approximately 1.2 TB of prepared data and TPU-pod training. An inference example does not reproduce that training process. [1]
Match observations and history
Base 1.5's model card describes primary RGB images at 256 by 256 and wrist images at 128 by 128, with batch and history dimensions. It allows subsets of the listed observation and task keys and a history window of up to two timesteps. The environment adapter guide requires observation keys to match the keys used in the training configuration; absent keys are padded. [6] [8]
History padding is a different concern from a missing camera. The README explains that timestep_pad_mask marks which history timesteps exist, including the absent previous observation at the start of a trajectory. The per-modality mask handles missing observation elements. Using these mechanisms deliberately avoids presenting missing information as a real measurement. [1]
The offline notebook adds batch and history dimensions explicitly, creates a language task through the model helper and samples an action chunk. It also passes action unnormalization statistics for the Bridge dataset. Those statistics belong to that example's data conventions; copying them into an unrelated robot integration does not establish the correct physical action interpretation. [5]
An action chunk is not a complete plan
Base 1.5 predicts four seven-dimensional actions at a time. The README offers several execution choices: execute the chunk, execute only its first action before resampling, or use temporal ensembling. None of these turns a single output into a full task trajectory. Choose the action execution scheme together with the robot interface and evaluation configuration. [1] [8]
The fine-tuned ALOHA simulation example uses its own environment, checkpoint, normalization statistics and history/action wrappers. It also specifies separate simulator dependencies and an external ACT checkout. It is evidence of an integration example, not a universal adapter or proof of performance on another robot. No rollout, real-robot trial or inference-speed measurement was performed for this profile. [7]
What we checked
Source research only; no installation, physical build or performance test performed. No affiliate links added.
Not yet established
- Requirements: operating system, host computer
- A step-by-step setup summary
- Version-specific hardware compatibility (ecosystem tags alone do not establish it)
Updates
Source files and licenses
Available project sources support research coverage; complete build and commercial qualification remain separate.
| Layer | Availability / license |
|---|---|
| software | available · MIT [3] |
| model weights | available · MIT Pinned Base1.5 model card reports MIT. Checkpoint payload not downloaded or executed; this finding is scoped to this model-card revision. [1] [8] |
| datasets | available · License not fully assessed Upstream links the Open X-Embodiment collection and preparation script. Constituent dataset licenses and payloads were not individually verified; no uniform dataset license asserted. [1] [8] |
| documentation | available · License not fully assessed Examples and source documentation available; separate documentation-wide license not established. [1] [4] [5] |
Model rights do not settle dataset rights
The source repository carries an MIT license, and the inspected Base1.5 model card separately reports MIT. The training mixture comprises multiple datasets, and the source links to their collection and preparation tooling. This review has not verified each constituent dataset's terms or downloaded the checkpoint and dataset payloads. It therefore makes no blanket licensing or complete reproduction claim for the training corpus. [1] [3] [8]
References
Each source links to the exact revision we read. “Used for” shows where this page relies on it.
- S001 octo-models/octo: README.md
Revision, date & where cited
github.com · revision
5eaa5c6· Accessed Oct 1, 2026
Used for: Overview, Is this for you?, What you need, Installation & first result and 5 more - S002 octo-models/octo: commit 5eaa5c6
- S003 octo-models/octo: LICENSE
Revision, date & where cited
github.com · revision
5eaa5c6· Accessed Oct 1, 2026
Used for: Source files and licenses - S004 octo-models/octo: files
Revision, date & where cited
github.com · revision
5eaa5c6· archive file inventory · Accessed Oct 1, 2026
Used for: Installation & first result, Source files and licenses - S005 Octo v1.5: examples/01_inference_pretrained.ipynb
Revision, date & where cited
github.com · revision
5eaa5c6· Accessed Oct 1, 2026
Used for: Is this for you?, What you need, Installation & first result, Hardware, models & data and 3 more - S006 Octo v1.5: examples/envs/README.md
Revision, date & where cited
github.com · revision
5eaa5c6· Accessed Oct 1, 2026
Used for: Is this for you?, What you need, Hardware, models & data, Match observations and history - S007 Octo v1.5: examples/03_eval_finetuned.py
Revision, date & where cited
github.com · revision
5eaa5c6· Accessed Oct 1, 2026
Used for: An action chunk is not a complete plan - S008 Octo Base 1.5 checkpoint model card
Revision, date & where cited
octo-models / project contributors · huggingface.co · README.md · Accessed Oct 1, 2026
Used for: Overview, Is this for you?, Hardware, models & data, Select a versioned checkpoint and 3 more
Correction guidance · Discussion is not enabled · Profile format 0.6.0-draft.1