allenai/Olmo-3.1-32B-Think
Text Generation
•
32B
•
Updated
•
580
•
19
Olmo 3.1
The latest members of the Olmo 3 family: another 3 weeks of RL for 32B Think, the 32B Instruct model, large post-training research datasets...
Note 📈 Scaling RL to make our latest model.
-
allenai/Olmo-3.1-32B-Instruct-SFT
32B • Updated • 445 • 3 -
allenai/Olmo-3.1-32B-Instruct-DPO
Text Generation • 32B • Updated • 723 • 2
allenai/Olmo-3.1-32B-Instruct
Text Generation • 32B • Updated • 155 • 8Note 💨 Our best model yet for chat & sensitive tasks.
allenai/Olmo-3.1-7B-RL-Zero-Math
Text Generation • 528k • Updated • 7 • 3Note 🧮 Improved RL Zero performance & more training steps!
allenai/Olmo-3.1-7B-RL-Zero-Code
Text Generation • 528k • Updated • 7 • 4Note 💻 Improved RL Zero performance & more training steps!
allenai/Dolci-Think-RL-7B-Completions-SFT
Viewer • Updated • 636k • 3 • 3Note Large datasets of completions used to filter prompts for our RL runs.
-
allenai/Dolci-Think-RL-7B-Completions-DPO
Viewer • Updated • 556k • 5 • 1
allenai/Dolci-DPO-Model-Response-Pool
Viewer • Updated • 71.2M • 44Note A very large set of completions across many models for preference tuning and reward modeling research.