Floor-weighted, not averaged. The weakest factor caps the level, because ownership is a conjunction. No factor is weak and both use-and-modify and data-control are strong, so the level is substantial; a moderate elsewhere keeps it short of full.

You substantially own OLMo: Apache-licensed, fully reproducible from open data and code, and self-hostable so nothing leaves your infrastructure. What holds it below full is capability - it is solid-in-class, not class-topping, and misuse is unbenchmarked with no companion guard model, so you supply the safety layer. Adopt it for any team that must prove what it runs - regulated, research, or EU-facing deployments where reproducibility and the Article 53 exemption matter - deploying the Instruct variant behind your own input/output guardrails.

Use and modify freelyCan you run, modify and adapt it with no gate and no field-of-use trap?

Strong

Apache-2.0 weights with unconditional commercial use, and full trainability - the OLMo-core training code and the Dolma corpus (ODC-BY) let you fine-tune, continue-pretrain or reproduce from scratch, no gate or field-of-use limit.

How this scores (AOI sub-dimensions)

Openness5/5how much is released - weights, data, code, licence - and how freelyEvery one of the six openness components is Open: Apache-2.0 weights, the full Dolma training dataset, runnable training code, released evaluation code and results, thorough technical reports, and an OSI-approved license.

Legal5/5how permissive and clean the licence is for real commercial useOSI-approved Apache-2.0 across weights, code and data; not a systemic-risk model; and - uniquely - the full training corpus is published, so a downstream deployer can assemble the EU AI Act training-content summary and technical documentation from released artifacts.

TransparencyDo you know what it is: weights, training, behaviour, and legible terms?

Strong

The registry's most auditable model: weights, training data, training code and evaluation are all open, so you can see not just the model but exactly how it was made.

How this scores (AOI sub-dimensions)

Provenance4/5how well we can trace and verify what went into the modelDistributed from the verified allenai org on Hugging Face in safetensors with checksums, and because training data, code and intermediate checkpoints are all public the pipeline is maximally auditable (checklist 6/8).

Governance4/5how accountable and well-documented the publisher isReputable, accountable US non-profit with a clear release cadence (OLMo 2 then OLMo 3) and full technical reports; short of 5 without a formally documented vulnerability-disclosure and deprecation policy.

Doesn't fail youIs it reliable and good enough for the job?

Moderate

Solid within its class (headline 80.8) rather than frontier-topping, and misuse is unbenchmarked with no companion guard model, so the safety layer is yours to supply.

How this scores (AOI sub-dimensions)

Performance3/5how capable it is relative to its classOn independent leaderboards OLMo is competitive within its size class but not frontier-leading on raw capability; its value proposition is transparency and reproducibility rather than class-topping benchmarks.

Operational4/5how practical it is to run, serve and maintain in productionStandard safetensors checkpoints run on mainstream stacks (vLLM, llama.cpp, Ollama, TGI, transformers) with documented hardware needs and community quants; short of a 5 mainly on the breadth of official quantizations and day-0 tooling versus the largest ecosystems.

Safety3/5whether misuse risks are evaluated and guardrails are providedInstruct variants are safety-tuned and documented, and withstand casual jailbreaks, but safety tuning is lighter than the large commercial labs and there is no companion guard model or broad independent red-team suite across CBRN/cyber domains.

Doesn't extract your dataDoes running it keep your knowledge and data yours?

Strong

Self-hosted, the open weights run on your own infrastructure under Apache-2.0 with no clawback or telemetry - your data and derived knowledge stay yours.

How this scores

Not a scored AOI dimension. For a self-hosted model, data-control is a structural property of running the weights yourself, strong by default unless the model phones home or the licence claws back rights. For a hosted API this factor is the retention + train-on-inputs + residency read, scored from the binding terms.

How the AOI score is computed

The seven dimensions above, each scored 0 to 5, weighted and summed to the 0 to 100 headline. The score is the analytical input behind the ownership verdict, not the verdict itself.

DimensionScoreWeightPoints

Openness5/50.1818.0

Provenance4/50.1612.8

Legal5/50.1616.0

Safety3/50.169.6

Performance3/50.148.4

Operational4/50.129.6

Governance4/50.086.4

HeadlineB · 80.8/100

Dossier coverageAssess 93%Implement 96%Use 61%Support 50%How complete our four-domain documentation is, a measure of our coverage, not of the model. Each domain links to its page.

Sources

Every rating traces to a primary document. Read means the text was verified; unverified means it is known to exist but has not yet been read.

DocumentWhat it grounds

Vendor announcementunverified2026-07-25

Ai2 presents OLMo as a fully open family releasing weights, training data, code and evaluation.

Model cardread2026-07-25

OLMo 2/3 checkpoints are hosted on the verified allenai org on Hugging Face in safetensors; the OLMo-2-1124-7B card states "The code and model are released under Apache 2.0" with Stage 1 OLMo-Mix-1124 (~3.9T) and Stage 2 Dolmino-Mix-1124 (843B).

Licenceread2026-07-25

The Dolma training dataset used for OLMo is publicly released under ODC-BY (Open Data Commons Attribution), transitioned to ODC-BY as of April 15, 2024, with users "also bound by any license agreements and terms of use of the original data sources."

Licenceread2026-07-25

OLMo code and weights are distributed under OSI-approved Apache-2.0, read verbatim from the repository LICENSE ("Apache License, Version 2.0, January 2004") and corroborated by the OLMo-2 model card.

Third-party analysisunverified2026-07-25

The Linux Foundation Model Openness Framework classifies OLMo among the most open (Open Science / Class I) model releases.

Model cardread2026-07-25

Canonical allenai repos publish per-file checksums; weights are not cryptographically signed.

Third-party analysisunverified2026-07-25

OLMo Instruct variants are safety-tuned but with lighter alignment coverage than large commercial labs.

Third-party analysisunverified2026-07-25

Independent leaderboards show OLMo competitive within its size class but not frontier-leading.

Documentationunverified2026-07-25

OLMo safetensors checkpoints load on mainstream serving stacks (vLLM, llama.cpp, Ollama, TGI, transformers).

Documentationread2026-07-25

OLMo training and evaluation code is publicly released on GitHub; the allenai/OLMo README (read this pass) is marked "out of date...no longer active" and lists released weights, both-stage checkpoints, training data, code and W&B logs under Apache-2.0.

Technical_reportread2026-07-25

Ai2's OLMo technical reports - "2 OLMo 2 Furious" (arXiv 2501.00656) and "Olmo 3" (arXiv 2512.13961) - document fully-released artifacts (weights, full training data, code, recipes, logs, thousands of intermediate checkpoints); Olmo 3 is "a family of state-of-the-art, fully-open language models at the 7B and 32B parameter scales."