Open research collective building open models, datasets and evaluation tools.
EleutherAI is a volunteer-born lab that builds open AI models and, in 2025, showed a competitive model can be trained on fully licensed data. Its evaluation tools are industry standard.
Focus
- Open source
- Transparency
- Data rights
Funding, as far as we know
Donations and compute grants (historically CoreWeave, Stability AI, Hugging Face, Google TRC) - https://www.eleuther.ai/about
What they've actually done
Released the Common Pile v0.1, an 8TB dataset of openly licensed and public-domain text, plus Comma models trained only on it
6 June 2025
Maintains the LM Evaluation Harness used by labs and safety institutes for benchmark reporting
ongoing
Released GPT-NeoX and Pythia open models enabling interpretability research
2022-2023