Robots Atlas>ROBOTS ATLAS
Olmo 3-Think
AI Models

Olmo 3-Think

3 (7B / 32B)
A fully open reasoning model from Ai2's Olmo 3 family, trained on multi-step problems spanning math, code and general reasoning.
✓ Active✓ Public access⚖ Open sourceLLMReasoning model
🏢Allen Institute for AIProducer
Access:DownloadAPIDeployment:💻 Local☁ Cloud

Overview

Olmo 3-Think is the reasoning variant of the Olmo 3 family developed by the Allen Institute for AI (Ai2). It is created by training the Olmo 3 Base model on multi-step problems spanning math, coding and general reasoning, generating explicit chains of thought.

Olmo models are fully open — Ai2 releases not only the weights but also the training data, code and recipe, making them one of the most transparent language-model families. Olmo 3-Think is available at 7B and 32B parameter scales; a newer Olmo 3.1-32B-Think has also been released.

The models target long-context reasoning and are published under an open license, with weights available on Hugging Face among others.

Classification
LLMReasoning model
Access & deployment
DownloadAPI
LocalCloud
Weights: Open source
Key parameters
✓ Tools · ✓ Fine-tuning
📥 Input: text

Technical specification

License
Apache 2.0
Features:✓ Tool use✓ Fine-tuning
Modalities
⬇ Input
text
⬆ Output
textcode

Sources and related pages