OLMo-core
Project record
OLMo-core is Ai2's Python library of building blocks for large-scale distributed training with PyTorch, used for its OLMo language models. The repository also holds the official training scripts for released OLMo 2 and Olmo 3 models and has beta support for text generation, including a chat demo. Version 3.0.0 was released on October 1, 2026.153
- Repository: Repository (external site: github.com)
- Documentation: Documentation (external site: olmo-core.readthedocs.io)
- License: LICENSE (Apache-2.0) (external site: github.com)
- Website: Package on PyPI (ai2-olmo-core) (external site: pypi.org)
- Release notes: v3.0.0 release notes (external site: github.com)
Availability and license
Overall availability
Source is public on GitHub, the package is on PyPI, and Ai2 publishes Docker images with its dependencies.14
Availability is separate from permission: read the license before using or redistributing.
Component reuse rights
- weights
- Unknown — no complete fact-level rights review
- code
- Reviewed qualifying license recorded — check scope and conditions
- data
- Unknown — no complete fact-level rights review
- documentation
- Unknown — no complete fact-level rights review
No complete system-rights review is recorded for this release.
Public materials checklist
| Item | Status | Notes and evidence |
|---|---|---|
| Source codeIs the source code publicly readable? | Public | Library source in src/olmo_core, tests in src/test, and official training scripts in src/scripts/official.1 |
| DocumentationIs user documentation published? | Public | The Read the Docs site has an overview, guides, examples, and an API reference. At review the site showed version 2.6.0, while the latest tagged release was 3.0.0.53 |
| InstallationAre installation instructions or packages publicly available? | Public | The README says to install PyTorch for the user's system first, then install from PyPI or from source; some features need optional dependencies, and the published Docker images include them.1 |
| Supported platformsAre supported operating systems or hardware documented? | Partial | The package requires Python 3.10 or later. The README says the Docker images are tested on Ai2's in-house H100 clusters and may not work with other hardware or driver and CUDA versions; no general platform list is given.41 |
| Release statusAre versioned releases published? | Public | Tagged GitHub releases with release notes and matching PyPI versions; 3.0.0 was released on 2026-10-01.34 |
What it is useful for
Run and use notes
- The 3.0.0 release notes say it removed the olmo_core.model_ladder API and its training scripts, so existing ladder configurations need an earlier revision.3
- The official training scripts can also be launched with OLMo-core's Beaker launch tool, which the README says requires access to Beaker; otherwise they run with torchrun.1
Organization context
U.S. eligibility
Eligible · basis: U.S.-governed project
The repository is in Ai2's allenai GitHub organization, the package is published on PyPI as ai2-olmo-core, and the documentation says OLMo-core is developed and maintained at the Allen Institute for Artificial Intelligence (AI2). Ai2 describes itself as a Seattle-based nonprofit AI research institute founded in 2014.1456
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.