Pythia
Family overview — summarizes releases; licenses and availability belong to each release.
Maintained by EleutherAI12
Pythia is EleutherAI's suite of language models built for research on how models learn during training and change with scale. The main suite has eight sizes from 70M to 12B parameters, each trained once on the Pile and once on a deduplicated Pile, on the same data in the same order, with 154 checkpoints per model. The current suite was retrained and re-released in April 2023; 14M and 31M models and additional random-seed runs were added later.123
- Repository: Pythia repository (external site: github.com)
- Paper: Pythia paper (arXiv 2304.01373) (external site: arxiv.org)
- Model hub: EleutherAI on Hugging Face (external site: huggingface.co)
Releases assessed
| Release | Weights | Tier (USASI rubric v0.1) | License | Released |
|---|---|---|---|---|
| Pythia 12B | Weights: Public | Model-disclosure tier (USASI rubric v0.1): Fully open | Apache-2.0, Apache-2.0 | Apr 3, 2023 |
What it is useful for
Organization context
Provenance and derivatives
U.S. eligibility
Eligible · basis: U.S. nonprofit or lab
The models are developed and released by EleutherAI, which states it incorporated as a non-profit research institute in early 2023; the IRS exempt-organization extract for the District of Columbia lists EleutherAI Institute in Washington, D.C., as a 501(c)(3) organization. Eligibility of each release is assessed on the release record.245
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.