Check weight access
Confirm that the model weights are actually available and identify the canonical source repository. Weight availability alone says nothing about the legal rights attached to those files.
OPEN-WEIGHT AI LICENSES · SOURCE-LINKED EVIDENCE
Understand which licenses are declared across 811 field-verified open-weight model records, how commercial-use status is classified, and where to inspect the exact source evidence before deployment.
Open-weight describes access to model weights, not one legal regime. Some repositories use permissive licenses such as Apache-2.0 or MIT; others use publisher-specific agreements, research terms, non-commercial licenses or custom conditions. A model can therefore be open-weight without being open source, unrestricted or automatically approved for commercial use.
Open Model Weights separates weight availability, the declared license, the commercial-use classification and the source evidence. That separation is essential for reliable model selection.
HOW TO READ LICENSE EVIDENCE
A familiar license name is useful, but it is not the whole decision. Open Model Weights keeps the evidence layers separate so the source can be inspected directly.
Confirm that the model weights are actually available and identify the canonical source repository. Weight availability alone says nothing about the legal rights attached to those files.
Identify the license named by the publisher or repository. Standard identifiers such as Apache-2.0 or MIT are normalized, while custom model licenses remain distinct instead of being forced into a generic category.
Review whether the checked terms are classified as allowed, conditional, revenue-conditioned, separate-license-required, not allowed or unknown. Unknown remains unknown when the source is insufficient.
Licenses can change between model versions or repositories. Use the model-level source links and verification date before relying on a classification for production, redistribution or commercial deployment.
If the intended use is commercial, regulated, redistributed to customers or embedded in a product, inspect the exact model-specific license file rather than relying only on the family name or a community summary.
COMMERCIAL-USE STATUS GUIDE
These labels summarize source evidence for discovery. They do not replace the publisher’s license text or legal review.
Commercial use is classified as allowed under the identified license terms.
Commercial use is classified as allowed, subject to additional terms stated by the publisher.
The checked source terms are classified as not allowing commercial use.
The checked source evidence is not sufficient for a confident commercial-use classification.
LICENSE GROUPS
Groups are normalized case-insensitively for standard license identifiers, so equivalent labels such as Apache-2.0 and apache-2.0 are not presented as separate licenses.
Apache License 2.0 is a permissive software license. Commercial use is generally permitted, while notice, attribution and patent-related terms still apply. Always inspect the exact repository license file.
Common developers: Qwen, Mistral AI, Google, Comfy-Org
The MIT License is permissive and generally permits commercial use, modification and redistribution while requiring preservation of the copyright and license notice.
Common developers: Microsoft, DeepSeek, BAAI, Z.ai
The source does not map cleanly to a standard license identifier. Inspect the model-level evidence before relying on any use right.
Common developers: NVIDIA, Meta, Qwen, Comfy-Org
The checked standard metadata did not declare a license for these records. Missing license evidence remains unknown rather than being treated as permission.
Common developers: OpenAI, Sentence Transformers, NVIDIA, DeepBeepMeep
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: NVIDIA, Poolside
The standard BY-NC family limits use to non-commercial purposes unless separate permission applies. Model-specific source terms remain authoritative.
Common developers: jinaai, Meta, Mistral AI, MahmoudAshraf
This is a publisher-specific model license. Open Model Weights reports the source-derived commercial-use classification but does not collapse custom conditions into a generic open-source label.
Common developers: Google, HauhauCS, GoTo / AI Singapore
This is a publisher-specific model license. Open Model Weights reports the source-derived commercial-use classification but does not collapse custom conditions into a generic open-source label.
Common developers: Meta
CC BY 4.0 permits sharing and adaptation, including commercial use, subject to attribution and the license conditions.
Common developers: pyannote, mlx-community, cardiffnlp, NVIDIA
The standard BY-NC family limits use to non-commercial purposes unless separate permission applies. Model-specific source terms remain authoritative.
Common developers: ProsusAI, audeering, Microsoft, E-MIMIC
OpenRAIL-family licenses can include use-based restrictions in addition to ordinary redistribution terms. The exact source text should be reviewed for the specific model.
Common developers: Comfy-Org, stable-diffusion-v1-5, CompVis, RunDiffusion
This is a publisher-specific model license. Open Model Weights reports the source-derived commercial-use classification but does not collapse custom conditions into a generic open-source label.
Common developers: Meta
BSD 3-Clause is permissive and generally allows commercial use and redistribution subject to its notice and non-endorsement conditions.
Common developers: Salesforce, MIT
This is a publisher-specific model license. Open Model Weights reports the source-derived commercial-use classification but does not collapse custom conditions into a generic open-source label.
Common developers: Meta, llava-hf, codellama
This is a publisher-specific model license. Open Model Weights reports the source-derived commercial-use classification but does not collapse custom conditions into a generic open-source label.
Common developers: Meta, NousResearch
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Mistral AI
OpenRAIL-family licenses can include use-based restrictions in addition to ordinary redistribution terms. The exact source text should be reviewed for the specific model.
Common developers: datalab-to, Datalab
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Preferred Networks
OpenRAIL-family licenses can include use-based restrictions in addition to ordinary redistribution terms. The exact source text should be reviewed for the specific model.
Common developers: Stability AI, SG161222, cagliostrolab
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: apple
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: BigScience
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: nlpaueb, airesearch
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Liquid AI
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: MiniMax
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Meta
This is not a standard permissive software license. Research or evaluation rights may differ from commercial rights; inspect the model-specific source license.
Common developers: Qwen
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: stepfun-ai, StepFun AI
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Bingsu
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Black Forest Labs
This is a publisher-specific model license. Open Model Weights reports the source-derived commercial-use classification but does not collapse custom conditions into a generic open-source label.
Common developers: Meta
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: MiniMax
This license group is reported from publisher or repository evidence. Rights and restrictions can vary by model, version and source text.
Common developers: Mistral AI
No license groups match the current filters.
LICENSE FAQ
Short answers are grounded in the distinction between weight availability, declared license terms and source-linked commercial-use evidence.
No. Open-weight means that model weights are available under some access terms. The license can still be custom, non-commercial, research-only, source-available or otherwise restricted. Open Model Weights reports the declared license separately from weight availability.
Sometimes. Commercial-use rights depend on the exact license and model-specific terms. This page classifies the source evidence as allowed, allowed with conditions, revenue-conditioned, separate-license-required, not allowed or unknown. It is evidence reporting, not legal advice.
Standard permissive licenses such as Apache License 2.0, MIT and BSD 3-Clause generally permit commercial use, modification and redistribution subject to their stated conditions. Model-specific files, additional terms and the exact repository version still matter.
The license field records the publisher-declared or source-observed license. Commercial-use status is a separate Open Model Weights classification of what the checked source terms appear to permit. Keeping the two separate avoids treating a familiar license label as a substitute for model-level evidence.
Publishers can attach different versions, additional terms, revenue conditions or model-specific restrictions. Open Model Weights keeps the model-level source evidence authoritative rather than assuming that a familiar label always has identical terms.
Unknown means the checked source did not provide enough evidence for a confident classification. Open Model Weights does not turn missing evidence into permission or prohibition.
The registry checks the source model repository, repository license files and declared metadata. Each model page links back to its evidence so the exact source can be inspected.
No. Open Model Weights is an evidence and discovery layer. For production, redistribution or commercial deployment, review the exact license attached to the model version you intend to use and obtain legal advice where appropriate.
Open Model Weights is an evidence and discovery layer, not a law firm. A classification such as “allowed” or “allowed with conditions” is a source-derived summary, not legal advice. For production or commercial deployment, inspect the exact license attached to the specific model version and obtain legal review where appropriate.
Read the verification methodology →Read the source policy →