OLMo 2 32B
Open Weights
Allen Institute for AI
Ai2's fully open 32B model (March 2025) - the first fully open model (all data, code, weights, and training details public) to outperform GPT-3.5 Turbo and GPT-4o mini on a suite of multi-skill academic benchmarks. Trained on up to 6T tokens and post-trained with Tulu 3.1.
Strengths
- Fully open: all data, code, weights, and training details public
- First fully open model to beat GPT-3.5 Turbo and GPT-4o mini
- Trained on up to 6T tokens, post-trained with Tulu 3.1
- Reproducible and auditable end to end
- Permissive Apache 2.0 license
Caveats
- Native 4K context (8K via RoPE scaling) - short vs frontier models
- Text-only (no vision or audio input)
- 32B trails the largest frontier models overall
- Vendor
- Allen Institute for AI
- Context Window
- 4,096 tokens
- License
- Apache-2.0
- Release Date
- 2025-03-13
- Modalities
- text
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Fully open under Apache 2.0 - all data, code, weights, and details public; free to self-host.