← back to the directory
Model Inference & Serving

OLMo 2

1Ai2's 2fully 3open 4reproducible 5language 6model

the six Ws · specification

W1 Who

The Allen Institute for AI (Ai2) developed the OLMo 2 model family.

W2 What

OLMo 2 is a fully open LLM family (1B, 7B, 13B, 32B) released with weights, training code, data, and logs.

W3 Where

Available on GitHub at allenai/OLMo and on Hugging Face under the allenai organization.

W4 When

OLMo 2 released in November 2024, with the 32B variant following in March 2025.

W5 Why

Built to give researchers complete transparency into pretraining data and methodology, not just final weights.

W6 With

Trainable and runnable via the OLMo-core codebase or Hugging Face Transformers.

W7 Watch

Apache 2.0 license, fully open weights, training data, and code with no gating, no credentials required, self-hosted; maintained by the Allen Institute for AI.

alternatives

works with

Allen AIFully openLLMApache-2.0Open training data

for agents & scripts

Reading this as a machine? Query it directly.

Search is open JSON - no key. Report telemetry after using a tool and it feeds that tool’s Proof Score. Or speak MCP to /mcp and discover tools mid-loop.