LLM Reference
EleutherAI

EleutherAI

12 models across 2 families · Latest: Llemma 34B (2023-09)

Championing open-source AI for everyone

ClassificationJSON / Tool useNon-Profit

EleutherAI's portfolio covers 12 active models across 2 current families, spanning classification and json / tool use. Open a model detail page to compare provider routes and sourced benchmarks.

Covers 2 workload areas across 12 active tracked models; last verified 2026-05-19.

Use it for

  • Teams evaluating classification and json / tool use across this lab's releases
  • Comparing model families before committing to a flagship
  • Migration and pricing follow-ups across 12 tracked models

Do not use it for

  • Choosing a hosting provider without opening a model page for price ladders

Active models

12

Current models from this lab, excluding deprecated ones

Active families

2

Current model families from this lab

Open catalog

12 open

10 open source / 2 open weights

Lowest output price

$0.200 /1M

Cheapest tracked output across active models, per 1M tokens

Latest dated release

2023-09-26

Llemma 34B

Freshness

2026-05-19

Researched 60d ago

aging

Information

Founded2020
New York, New York, United States

Release cadence

Showing 5 recent dated releases (full timeline below). Latest: Llemma 34B (2023-09-26).

Where this lab wins

  • Classification: 1 tracked model with MMLU-class moderation/safety coverage.
  • JSON/tool-use: 1 tracked model with BFCL / Nexus strict-JSON routing coverage.

Flagship quality / price signal

Flagship: Llemma 34B (best sourced coding quality-per-dollar in this portfolio).

Quality-per-dollar unavailable for this flagship — benchmark coverage or output token pricing is still missing.

EleutherAI is an American AI research organization founded in 2020. Championing open-source AI for everyone. EleutherAI ships 2 model families totaling 12 models, with the most recent release Llemma 34B in 2023-09. Notable families include Llemma and Pythia. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are added. Researchers and evaluators can scan. View official API endpoints, benchmark performance, and coding/agent fit for every EleutherAI model.

About

EleutherAI's journey began as a modest Discord server where discussions centered around the newly released GPT-3 model. Founded in July 2020 by Connor Leahy, Sid Black, and Leo Gao, it quickly evolved into a pioneering non-profit AI research lab. With a primary focus on the interpretability and alignment of large language models (LLMs), EleutherAI has made significant strides in making advanced AI technologies more accessible, fueling the open-source AI community's growth and impact. The organization has gained recognition for developing and releasing several influential models, including GPT-J, GPT-NeoX, BLOOM, VQGAN-CLIP, Stable Diffusion, and OpenFold. These models have seen widespread adoption, with over 25 million downloads, and have facilitated research across diverse domains such as interpretability, ethics, and training dynamics. A noteworthy contribution to the field is The Pile, a vast and diverse dataset of 825 GiB of text and code designed for training large language models. This dataset has been instrumental in advancing the capabilities of open-source LLMs. Another innovative aspect of EleutherAI's work is its dedication to open science and collaboration. By operating primarily through a public Discord server, EleutherAI fosters a unique collaborative environment where researchers, volunteers, and external collaborators from around the world can contribute and innovate together. This model has been pivotal in EleutherAI's ability to develop and release impactful open-source AI models and datasets, embodying a commitment to open access that challenges the dominance of large technology firms. EleutherAI actively sponsors AI research and educational efforts globally, striving to broaden technology access and ensure that future AI advancements are not confined to a few powerful entities. Despite being listed as a for-profit company on Crunchbase, EleutherAI's own communication and reliable sources consistently refer to it as a non-profit organization. This discrepancy raises questions about financial structure transparency. They receive support from prominent entities such as Google TRC, Open Philanthropy, Mozilla Foundation, Omidyar Network, Stability AI, Lambda Labs, and Nat Friedman, though detailed insights into their financial intricacies remain limited. EleutherAI remains a dedicated advocate for open-source contributions, continuously pushing the envelope in AI research and development. Their unique approach and impactful contributions to the AI field reflect their mission to democratize access to AI innovations globally.

Featured models

ModelReleasedContextInput price ($/1M)Output price ($/1M)LicenseOpenness
Llemma 34B2023-09-264k--Llama 2 CommunityOpen weights
Llemma 7B2023-09-264k--Llama 2 CommunityOpen weights
Pythia 1.4B2023-05-312k--Apache 2.0Open source

Model families

Recent releases

  1. Llemma 34B- 2023-09-26
  2. Llemma 7B- 2023-09-26
  3. Pythia 1.4B- 2023-05-31
  4. Pythia 410M- 2023-05-31
  5. Pythia 12B- 2023-05-31

FAQ

Who founded EleutherAI and when?

EleutherAI was founded in 2020 and is associated with New York, New York, United States.

What models has EleutherAI released?

EleutherAI ships 12 models across 2 families: Llemma and Pythia.

Is EleutherAI's technology open source?

All tracked EleutherAI models are open-weight or open-source.

Where is EleutherAI headquartered?

EleutherAI is headquartered in New York, New York, United States.

What is EleutherAI known for?

Championing open-source AI for everyone. Its most prominent tracked family is Llemma.

How can I access EleutherAI's models?

EleutherAI's models are available via Fireworks AI.

Explore related pages

Last reviewed: 2026-05-19. Data sourced from public lab announcements and provider documentation.