Overview
Olmo 3 continues Ai2’s practice of opening the whole pipeline, not just the weights. The models are pre-trained on the Dolma 3 dataset and post-trained on the Dolci datasets, and Ai2 calls Olmo 3 Think 32B the first fully open 32B thinking model. Olmo 3.1 updates followed in December 2025.
Highlights
- 7B and 32B sizes
- Base, Instruct and Think variants
- Training data, code and intermediate checkpoints published
- 65,536-token context window
- Apache 2.0 license
Typical uses
Research on training and data, transparent assistants and fine-tuning.
Things to consider
- Ai2 lists English as the primary language.
- Intended for research and education under Ai2’s responsible-use guidelines.
- Download sizes refer to the default quantized versions in the Ollama library; full-precision weights are larger.
Specifications are taken from the model card on Hugging Face, the Ollama library and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.