Olmo 3
Team Olmo, Allyson Ettinger, Amanda Bertsch, Bailey Kuehl, David Graham, David Heineman, Dirk Groeneveld, Faeze Brahman, Finbarr Timbers, Hamish Ivison, Jacob Morrison, Jake Poznanski, Kyle Lo, Luca Soldaini, Matt Jordan, Mayee Chen, Michael Noukhovitch, Nathan Lambert, Pete Walsh, Pradeep Dasigi, Robert Berry, Saumya Malik
2025-12-17
Summary
This paper introduces Olmo 3, a new set of powerful language models that are completely open to the public, coming in two sizes: 7 billion and 32 billion parameters.
What's the problem?
Existing large language models often aren't fully transparent – it's hard to know exactly how they were built, what data they were trained on, or even access the models themselves. This limits research and makes it difficult for others to build upon existing work. There's also a need for open-source models that can perform complex tasks like reasoning, coding, and having natural conversations.
What's the solution?
The creators of this paper built Olmo 3 from the ground up and released *everything* involved in its creation. This includes the model's code, all the data used for training, and even checkpoints showing its progress. Their most powerful model, Olmo 3 Think 32B, is designed to excel at challenging tasks requiring thought and problem-solving. They essentially provided a complete blueprint for building a high-performing language model.
Why it matters?
This work is important because it promotes open science and allows anyone to study, use, and improve upon a state-of-the-art language model. By releasing all the details, it fosters collaboration and accelerates progress in the field of artificial intelligence, making powerful AI technology more accessible to everyone.
Abstract
We introduce Olmo 3, a family of state-of-the-art, fully-open language models at the 7B and 32B parameter scales. Olmo 3 model construction targets long-context reasoning, function calling, coding, instruction following, general chat, and knowledge recall. This release includes the entire model flow, i.e., the full lifecycle of the family of models, including every stage, checkpoint, data point, and dependency used to build it. Our flagship model, Olmo 3 Think 32B, is the strongest fully-open thinking model released to-date.