EleutherAI
eleuther.ai
Nonprofit AI research collective that builds and openly releases frontier language models, evaluation frameworks, and interpretability research.
Why it is useful
EleutherAI produced GPT-Neo, GPT-J, and the Pile dataset before open-source LLMs became mainstream. They created the LM Evaluation Harness, the standard benchmarking tool used to compare models across the field. Following their research gives a more rigorous perspective on model capabilities than most commercial sources.