OLMo 2
verifiedFreeallenai.org
Ai2's family of fully open language models with weights, code, and training data released, built for transparent LLM research and building.
What it does
Olmo is Ai2's family of fully open language models. The current Olmo 3 lineup spans base, reasoning ('Think'), and instruct variants at 7B and 32B, released with open weights, code, training data, and the full model flow so researchers can inspect and customize every stage.
Core features
Fully open weights, code, and training data
Base, Think (reasoning), and Instruct variants
7B and 32B model sizes
Open model flow across all training stages
Open-source training/eval tools (OlmoCore, OLMES)
OlmoTrace to trace outputs to training data
Best for
→Researching language-model training and behavior
→Building and fine-tuning open models
→Machine-unlearning and clinical-NLP research
→Deploying transparent open LLMs
Toolspool rankingby global site rank