Ai2 Releases Olmo 3: A Fully Open 32B 'Thinking' Model
The Allen Institute challenges closed labs with a radically transparent reasoning model.
The Allen Institute for AI (Ai2) challenged the industry's trend toward secrecy, with the release of the Olmo 3 family of models. The release includes a 32-billion parameter model capable of "thinking" and complex reasoning, designed to compete with proprietary systems like OpenAI’s o1.
Radical Transparency
Unlike "open weights" models (like Llama) which hide their training data, Olmo 3 is "fully open." Ai2 has released: * The full model weights. * The complete Dolma 3 training dataset. * The full training code, logs, and intermediate checkpoints.
The "Thinking" Capability
This is the first fully open-source model to feature …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.