Ai2 Open Models: Model List and Licenses

About Ai2

Ai2 (Allen Institute for AI) is a US non-profit AI research institute. It publishes 970 models on its allenai organization on Hugging Face and is known for building fully open language models for research.

Its model cards describe the Olmo family as “Open language models designed to enable the science of language models". It releases not only finished models but also many intermediate training stages and research derivatives.

What It Releases

  • Olmo 3 and 3.1 (November to December 2025): 7B and 32B, in Instruct versions and Think versions that strengthen math and coding reasoning with long chains of thought.
  • OLMoE: an MoE with 7B total and 1B active parameters.
  • olmOCR 2 (October 2025): a document-reading model fine-tuned from Qwen2.5-VL-7B-Instruct; the FP8 version is recommended for practical use.
  • Molmo2: models that also read images and video. Molmo2-ER (May 2026, 4B, built on Qwen3-4B and SigLIP2) specializes in embodied perception and serves as the backbone of a robot action model.
  • Research models (2026): tmax, agent-training research models based on Qwen3.5-9B, MoE research checkpoints and more.

Licenses

The main models, including Olmo, OLMoE, olmOCR and Molmo2, are Apache-2.0. Some models use a custom license (other). Read each model’s license text before use. This summary is not legal advice.

Running Them Locally

  • Official GGUF builds exist for some models, such as OLMo 2, OLMoE and olmOCR (as of 2026-09-26).
  • Most models are research models, and many in the table have no stated task. For trying general conversation locally, the Olmo 3 7B is an easy candidate.

Sources: model cards of allenai/Olmo-3.1-32B-Think, allenai/olmOCR-2-7B-1025 and allenai/Molmo2-ER; the model list of the allenai organization on Hugging Face (all as of 2026-09-26).

Our Coverage and Data

Local Model Watch has not published an article on Ai2’s models yet. The model list below is collected by code from Hugging Face every day. Part of our publisher index; model families link to their family pages.

At a Glance

Item Value
Hugging Face organization allenai
Models on Hugging Face 970
Articles on its own models 0
Articles on derived models 0

Recent Models on Hugging Face

The publisher lists 970 models in total on Hugging Face. Below are up to 15 of the most recent generative models (text, image, video, audio and so on), the scope of this site, per organization. Classifiers, feature extractors, research components and repositories without a model card are omitted, and the publisher’s own quantized builds and helper variants are folded into the row of the original model. Parameter counts are left blank for repositories that pack weights into integer types, where Hugging Face’s count is not the real parameter count.

→ Scroll horizontally to see all columns

Released Model Task Parameters License Official variants
2026-06-22 allenai/tmax-27b — 3M apache-2.0 —
2026-06-19 allenai/qwen35-9b-openthoughts — 9.0B apache-2.0 —
2026-06-19 allenai/qwen35-9b-terminaltraj — 9.0B apache-2.0 —
2026-06-19 allenai/qwen35-9b-cli-gym — 9.0B apache-2.0 —
2026-06-19 allenai/qwen35-9b-swesmith — 9.0B apache-2.0 —
2026-06-19 allenai/qwen35-9b-termigen — 9.0B apache-2.0 —
2026-06-19 allenai/qwen35-9b-endless — 9.0B apache-2.0 —
2026-06-18 allenai/tmax-sft-8b — 8.2B apache-2.0 —
2026-06-17 allenai/tmax-8b — 8.2B apache-2.0 —
2026-06-17 allenai/tmax-4b — 4.2B apache-2.0 —
2026-06-17 allenai/tmax-2b — 1.9B apache-2.0 —
2026-06-17 allenai/tmax-9b — 9.0B apache-2.0 —
2026-06-16 allenai/MolmoMotion-4B-H3-F30 vision-language 4.9B apache-2.0 —
2026-06-16 allenai/MolmoMotion-4B-H1-F32 vision-language 4.9B apache-2.0 —
2026-05-09 allenai/EMO text generation 13.6B other —

From the Hugging Face API. Last changed 2026-09-27 (JST). Collected by code.

Last updated 2026-09-27 (JST). The explanation at the top of this page was written with the help of AI from the primary sources it cites. The tables and lists under “Our Coverage and Data" are assembled by code from our article log and the Hugging Face API.