The field guide
Compare two releases
Two releases can both be called open and still leave you with very different choices. Pick a question. Look at the evidence.
What does
open let you do?
Release A
Highly usable open-weight model with Apache 2.0 weights but incomplete training openness.
Profile, scope & sources ↗Release B
Open research system: the reference case for end-to-end openness.
Profile, scope & sources ↗OpenAI gpt-oss (120b / 20b)
Transparency
The dataset is described at a high level along with architecture, tokenizer, post-training approach, and safety evaluation, but no reproducible training corpus or recipe is published.
OLMo 2 / OLMo 3 / OLMo 3.1
Transparency
The whole model flow is documented: Dolma and Dolci datasets, the OLMo-core training framework, the open-instruct post-training pipeline, data-cleaning and decontamination tooling, the OLMES evaluation harness, OlmoTrace for tracing outputs back to training data, and stage-level checkpoints for the Instruct, Think, and RL Zero paths.
The book’s qualitative assessments · snapshot 2026-11-15. Scores describe the named artifacts, not intelligence or OSAID certification. Governance is shown separately from tier assignment.
Explore all 91 profiles