Allen Institute for AI (Ai2) model releases

Ai2's fully open thinking model with public weights, code, data, checkpoints, and training details across the OLMo 3 pipeline.

A fully open model — weights, data, and training code all public — and the first such to beat GPT-3.5 / GPT-4o mini.

Ai2's post-trained open instruction model line, scaling the Tulu recipe to Llama 3.1 405B.

Open multimodal model family trained for strong image understanding, pointing, and visual grounding.

Fully open sparse MoE model with 7B total and about 1B active parameters.

Ai2's first fully open language model release, including weights, training data, code, logs, and intermediate checkpoints.