<ul><li>The Allen Institute for AI (AI2) has developed OLMo 2, an open-source family of language models.</li><li>OLMo 2 consists of 7 billion (7B) and 13 billion (13B) parameter configurations and was trained on up to 5 trillion tokens.</li><li>The models bridge the performance gap with proprietary systems and outperform their predecessors as well as competitors.</li><li>Advancements in training stability, staged training, evaluation frameworks, and dataset diversity contribute to OLMo 2's success.</li></ul>

The Allen Institute for AI (AI2) Releases OLMo 2: A New Family of Open-Sourced 7B and 13B Language Models Trained on up to 5T Tokens

Discover more