OLMo 3 Instruct

weights

open(Apache-2.0)

data

open(Dolma 3 pretraining + open post-training mixtures)

code

open(OLMo + open-instruct)

checkpoints

open

license

Apache-2.0(OSI)

A fully open instruct line: Apache-2.0 weights, open pretraining and post-training data, open training code and open checkpoints. The line now extends to an Olmo 3.1 update shipping 32B Think and Instruct checkpoints on the same open recipe, which does not change the class.

• https://huggingface.co/allenai/Olmo-3-7B-Instruct recorded 2026-08-13

  `license:apache-2.0` in the repo tags and `"gated":false` in the embedded repo state, so the 7B Instruct weights download without a barrier

• https://allenai.org/blog/olmo3 recorded 2026-08-13

  the model flow is published in full - "every stage, checkpoint, dataset, and dependency required to create and modify it"; Models / Tech Report / Code / API / Demo links, with Code pointing at github.com/allenai/open-instruct; the 12/12 update adds Olmo 3.1 Instruct 32B on the same recipe

Solid research and practitioner reach for a fully open instruct line, below the Qwen and Llama instruct workhorses, and still ramping as it supersedes OLMo 2 Instruct. The declared artifact allenai/Olmo-3-7B-Instruct reads 450,053 Hugging Face downloads in the trailing 30 days, which bands at 100K-1M, level 3. The other declared artifact, allenai/Olmo-3-1025-32B-Instruct, answers HTTP 401 and contributes nothing to the sum; it could only raise the figure, never lower it.

• https://huggingface.co/allenai/Olmo-3-7B-Instruct recorded 2026-08-13

  450,053 downloads in the trailing 30 days for allenai/Olmo-3-7B-Instruct

The strongest fully open chat line, competitive with same-scale open-weight instruct models but below the 2026 frontier, which is where a 5 sits.

• https://allenai.org/blog/olmo3 recorded 2026-08-13

  "Olmo 3-Instruct ties or surpasses Qwen 2.5, Gemma 3, and Llama 3.1 in our evaluations, and competes with the Qwen 3 family at similar scale"